ArticlePDF Available

An Architecture for Integrated Online Analytical Mining

May 2011
Journal of Emerging Technologies in Web Intelligence 3(2)

DOI:10.4304/jetwi.3.2.74-99

Authors:

Muhammad Usman

Auckland University of Technology

Sohail Asghar

COMSATS University Islamabad

Online Analytical Processing (OLAP) technology is an essential element of the decision support system and permits decision makers to visualize huge operational data for quick, consistent, interactive and meaningful analysis. More recently, data mining techniques are also used together with OLAP to analyze large data sets which makes OLAP more useful and easier to apply in decision support systems. Several works in the past proved the likelihood and interest of integrating OLAP with data mining and as a result a new promising direction of Online Analytical Mining (OLAM) has emerged. In this paper, a variety of OLAM architectures in the literature were reviewed and the limitations in the previously reported work have been identified. Literature review reveals the fact that none of the previously reported OLAM architectures have integrated enhanced OLAP with data mining. We enhanced the performance of OLAP in terms of cube construction time and visualization by providing interactive visual exploration of data cube. Furthermore, automatic OLAP schema generation has never been introduced as a component of OLAM architecture. The aim of this paper is to propose an integrated OLAM architecture that not only overcomes the existing limitations but also extends the architecture by adding an automation layer for the schema generation. In the proposed work, hierarchical clustering has been used as the data mining technique and three types of schemas namely star, snowflake and galaxy were automatically generated. A prototype of automatic schema generation has been developed and schema generation algorithms have been provided. In addition to this, we implemented and deployed the proposed architecture. Validation has been done by performing experiments on real life data set. Experimental results prove that the proposed architecture improves the cube construction time, empowers interactive data visualization, automate schema generation, and enable targeted and focused analysis at the front-end. By integrating enhanced OLAP with data mining system a higher degree of development is achieved which makes significant advancement in the modern OLAM architectures.

Proposed architecture for integrated OLAM

…

ABBREVIATION OF CENSUS DATA CLUSTERS

…

Sequence of architectural implementation steps

…

OLAM ARCHITECUTE COMPARISON TABLE

…

Clustering parameter selection in HCE tool.

…

Figures - uploaded by Muhammad Usman

Content may be subject to copyright.

Content uploaded by Muhammad Usman

Content may be subject to copyright.

An Architecture for Integrated Online Analytical

Mining

Muhammad Usman

Department of Computer and Informataion Sciences, Auckland University of Technology, New Zealand

Email: usmanspak@gmail.com

Sohail Asghar

Center of Research in Data Engineering (CORDE),

Muhammad Ali Jinnah University, Islamabad, Pakistan

Email: sohail.asghar@jinnah.edu.pk

Abstract— Online Analytical Processing (OLAP) technology

is an essential element of the decision support system and

permits decision makers to visualize huge operational data

for quick, consistent, interactive and meaningful analysis.

More recently, data mining techniques are also used

together with OLAP to analyze large data sets which makes

OLAP more useful and easier to apply in decision support

systems. Several works in the past proved the likelihood and

interest of integrating OLAP with data mining and as a

result a new promising direction of Online Analytical

Mining (OLAM) has emerged. In this paper, a variety of

OLAM architectures in the literature were reviewed and the

limitations in the previously reported work have been

identified. Literature review reveals the fact that none of the

previously reported OLAM architectures have integrated

enhanced OLAP with data mining. We enhanced the

performance of OLAP in terms of cube construction time

and visualization by providing interactive visual exploration

of data cube. Furthermore, automatic OLAP schema

generation has never been introduced as a component of

OLAM architecture. The aim of this paper is to propose an

integrated OLAM architecture that not only overcomes the

existing limitations but also extends the architecture by

adding an automation layer for the schema generation. In

the proposed work, hierarchical clustering has been used as

the data mining technique and three types of schemas

namely star, snowflake and galaxy were automatically

generated. A prototype of automatic schema generation has

been developed and schema generation algorithms have

been provided. In addition to this, we implemented and

deployed the proposed architecture. Validation has been

done by performing experiments on real life data set.

Experimental results prove that the proposed architecture

improves the cube construction time, empowers interactive

data visualization, automate schema generation, and enable

targeted and focused analysis at the front-end. By

integrating enhanced OLAP with data mining system a

higher degree of development is achieved which makes

significant advancement in the modern OLAM

architectures.

Index Terms—Automatic schema, cube, decision support

system, data mining, data visualization, hierarchical

clustering, OLAP

I. INTRODUCTION

Online Analytical Processing (OLAP) technology refers

to a set of data analysis techniques to view the data from

all of the transactional systems in an interactive way in

order to support the decision-making process. According

to [1] and [2] OLAP systems have rapidly gained

momentum in both the academic and research

communities, mainly due to their allowance for quick,

multi-dimensional analysis capabilities. It plays a vital

role in data analysis to identify the key performance

indicators of business and other application domains. In

addition to this, it is an essential element of the decision

support system and permits decision makers to visualize

huge operational data for consistent and interactive

analysis. Furthermore, it is becoming more accepted due

to the recognition of the business value of exploring and

querying huge amounts of data sets in a multi-

dimensional way. In the last few years OLAP systems

are the predominant front-end tools used in data

warehousing environments and the OLAP system’s

market has developed rapidly. According to [3], the main

reasons for this rapid development are as follows;

i)These systems are being used by knowledge workers to

obtain quick answers of complex queries on their huge

operational data.

ii) Allow users to perform quick, efficient and interactive

analysis on organizational historical data.

iii) Specially designed with the aim of better supporting

the retrieval of higher level summary information from

detailed data.

iv) Offer better performance for aggregate queries

However, the growing complexity and volumes of the

data to be analyzed impose new requirements on OLAP

systems [4]. To fulfill these increasing requirements, a

number of OLAP enhancement techniques have been

proposed in the past [5,6,7,8,9,10,11,12,13,14,15,16].

Although some of the previously discussed research work

proved the enhancement of OLAP systems but none of

the work was intended towards focusing on the integrated

enhancement of OLAP performance and visualization

and there exist a strong need for it [17]. One of the main

contributions in this paper is the integrated enhancement

of OLAP which improves the performance in terms of

cube construction time and intensifies visualization of

cube data with the help of cube visualization API.

74 JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011

doi:10.4304/jetwi.3.2.74-99

Along with the integrated enhancement of OLAP, the

other major contribution is the combination of OLAP and

data mining. According to [18], data mining is a step in

the overall concept of knowledge discovery in databases

(KDD) and data mining techniques like Association [19],

Classification [20], Clustering [21] and Trend analysis

[22] can make OLAP more useful and easier to apply in

decision support systems. More recently, data mining

techniques have been applied in decision support

applications in order to detect patterns and to extract

knowledge form data. Several previous works [5, 7, 8 and

23] proved the likelihood and interest of coupling OLAP

and data mining. However, [24] introduced a new

concept of integrating OLAP and data mining and named

it as Online Analytical Mining (OLAM).

OLAM is a promising direction for mining knowledge

from multidimensional databases [23]. Significant works

in this area include [24, 25, 26, 27 and 28]. These authors

used different data mining techniques in their areas of

interest and produced better results through the concept

of OLAM. In this paper, a variety of OLAM architectures

were reviewed and their limitations have been identified

and highlighted. A major limitation revealed in literature

review is the fact that none of the previously proposed

architectures integrated enhanced OLAP with data

mining. In the proposed architecture, we integrated

enhanced OLAP with a data mining technique known as

hierarchical clustering.

Hierarchical clustering is one of the dominant data

mining techniques in the area of OLAM. In this paper

work, we have also used the hierarchical clustering

technique of data mining as a pre-processing step.

Previously, hierarchical clustering has been used by [29,

30, 31, 32, and 33] in conjunction with OLAP systems for

a number of reasons which include query optimization,

physical data organization, fast data access and reduced

storage cost. We intend to use the results of this

hierarchical clustering technique in the form of data

clusters to generate automatic schemas for enhanced

OLAP system.

As far as the automatic generation of OLAP schema is

concerned, [34, 35, 36, and 37] reported the work in this

area. To our knowledge none of these works used

hierarchically clustered data to generate OLAP schema.

Furthermore, authors have used ER diagrams or

conceptual graphical models to produce only one type of

schema which is mostly star schema. Another major

limitation in the automation work is that the tools and

techniques used for schema generation only draws or

identify the structure of schema and doesn’t populate the

actual data in the generated schema. In this paper work,

we are not only generating OLAP schema of three types

namely; star , snowflake and galaxy but our automated

schema builder tool provides automation in uploading the

data from clustered file to relevant portions of the

automatically generated schema.

On the basis of these observations and limitations, we

proposed an integrated OLAM architecture. The proposed

architecture, firstly, enhances the power of OLAP in

terms of its performance and visualization, secondly,

integrates the enhanced OLAP with data mining

technique of hierarchical clustering and lastly, extends

the existing OLAM architecture with the addition of

automatic schema generation layer. We have developed a

prototype of automatic schema generation layer and also

provided the schema generation algorithms. In addition to

this, we implemented and deployed the proposed

architecture.

Validation has been done by performing experiments on a

real life data set. Experimental results prove that the

proposed architecture improves the cube construction

time, empowers interactive data visualization, automate

schema generation, and enable targeted and focused

analysis at the front-end. By integrating enhanced OLAP

with data mining system a higher degree of development

is achieved which makes significant advancement in the

modern OLAM architectures. Finally, to the best of our

knowledge, the proposed integrated OLAM architecture

is superior to other OLAM architectures proposed prior

or in course of carrying out this research.

The rest of the paper is organized as follows. Section II

highlights the research contributions. Section III covers

the literature review, in section IV the integrated OLAM

architecture is proposed, Section V discusses the

architectural implementation, Section VI elaborates the

validation and evaluation of the proposed architecture.

Finally, in section VII the summary of the work and the

possible future direction from this research has been

discussed.

II. RESEARCH CONTRIBUTION

The title of this paper reflects the overall goal of the

work, which is to propose an integrated OLAM

architecture. Although many research areas in the field of

OLAM have reached the state of maturity, new

challenges arise when applying the OLAM architecture to

novel usage scenarios. In the following paper, we focus

on overcoming the deficiencies of the conventional

OLAM architecture in providing adequate and focused

analysis in decision support applications. We have

accomplished the following research contributions in this

paper:

• Improving OLAP performance in terms of cube

construction time and enhancing data

visualization through the use of visualization

API.

• Identifying a strong area of research that is the

integration of enhanced OLAP with data mining

to extend the OLAM architecture.

• Using hierarchical clustering as a pre-processing

step of data mining.

• Proposing the integrated OLAM architecture.

JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011 75

• Extending the OLAM architecture with the

addition of automatic schema generation layer.

• Supporting the automatic generation of three

types of OLAP schemas namely star, snowflake,

and galaxy.

• Automating the deployment of generated

schema in the database server along with the

population of data in the schema.

• Implementing the proposed architecture using

market based technological tools.

• Developing prototypes of automatic schema

generation layer and front-end targeted analysis.

• Performing experiments on real-life data sets

using the prototype to compare and evaluate the

results of the proposed architecture.

• Validating the experimental results and

highlighting the benefits attained.

III. LITERATURE REVIEW

In this section, literature review of integrated OLAM

technology is presented. The purpose of this literature

review is to study, analyze and identify the

limitations of the OLAM technology. As described

earlier, integrated OLAM technology is the combination

of enhanced OLAP and data mining technique with

automation in schema generation process. Additionally,

the focus of this work is the use of hierarchical

clustering to automate the process of OLAP schema

generation. This diversifies the scope of literature review

and comes under the larger umbrella of OLAM. Our

overview of related topics focuses on four major themes:

(1) OLAP enhancement, (2) integration of OLAP

enhancement techniques, (3), use of hierarchical

clustering with OLAP, and (4) automation in OLAP

schema generation.

A. OLAP Enhancement

In this section, we review various significant OLAP

enhancement techniques in the literature. The major focus

of this section is performance improvement and

visualization enhancement techniques. Firstly, we discuss

some of the performance improvement techniques

followed by the visualization enhancement techniques.

According to Goil et al. [5], parallel computing can be

used as a way to enhance the performance of OLAP. It

has been identified in the research work that typically

OLAP operations are responsible for running complex

queries on the underlying data, so these operations

increase data processing time. The introduction of high

performance parallel computers can lessen the

computational time by distributing the tasks among

various computer resources. The authors proposed an

algorithm for Cube construction on parallel computers

using distributed memory. The limitation of the work is

the usage of a small and simple data set for the

experiment. If the data set is large then a mismatch

in the order of the dimensions can occur, as the

multidimensional array doesn’t allow the memory’s use

of their sort method to be efficient.

Similarly, Papadias et al. [6], reported on the efficient

OLAP operations in Spatial Data Warehouses. Authors

identified the problem that the computation of spatial

data is expensive as online processing of such

complex spatial data is inapplicable. They suggested a

data structure, named aR-Tree, which combines a spatial

index with the materialization technique. The strength of

their work was the identification of the position of the

objects in space in which the groupings and the

hierarchies among the dimensions were unknown at

design time. The work done is limited only to Spatial

Data Warehouses and it also does not give any indication

of the OLAP storage types such as ROLAP and MOLAP

for Spatial data warehouses. The dataset used by the

authors was found to be insufficient for the acceptance of

this technique for all spatial cases.

In addition to the OLAP enhancement work, Asghar

et al. [7], proposed a functionality-enhancement

technique using Growing Self-organizing Neural

Networks. This technique proposes the integration of

data mining with OLAP by passing on mined data to the

OLAP engine for a more focused analysis and hence

adding intelligence to the OLAP system and thereby

helping its users in sharp analysis of data . The major

limitation is the manual schema generation of hierarchical

clusters for the OLAP engine to perform operations.

Furthermore, the data set has to be in a special format for

the generation of hierarchical clusters using Growing

Self-Organizing Map (GSOM) method. The enhancement

architecture provided by the authors does not satisfactory

answers for complex data sets. No indication was given

as to how the hierarchical clusters generated from the

dataset will be visualized.

To achieve better OLAP performance, a new

contribution in the performance improvement area by

associating OLAP and data mining was proposed.

Messaoud et al. [8] suggested an enhanced OLAP

operator based on the Agglomerative Hierarchical

Clustering (AHC). The main problem identified is

OLAP limitation of aggregation and summarization

of complex objects like text, images, sounds and

videos. The operator called Operator for Aggregation

by Clustering (OpAC) is able to provide significant

aggregates of facts referred to complex objects. The

strength of the work is the combination of OLAP and

data mining techniques as data mining can discover

knowledge from both simple and complex data. The

major limitation of the technique is that for n

individuals to classify, the AHC generates n

hierarchical partitions but it does not give guidance

as to a best partition to choose. Data miner has to

decide the number of clusters that corresponds to the

context and to the goal of his/her analysis.

So far we have discussed the performance improvement

techniques, now we shift our review focus to some of the

76 JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011

major OLAP visualization enhancement work. For the

purpose of visualization enhancement, Maniatis et al. [9],

recommended the Advanced Visualization of OLAP.

They suggested the Cube Presentation Model (CPM),

which can be naturally mapped with an advanced

visualization technique from the Human-Computer

interaction area, called Table Lens. They formalized the

simultaneous presence of more than one queries

which was done in two layers (Presentation Layer

and Logical Layer). The limitation of the work is that the

visualization technique used for the screens is restricted

only to one aspect of visualization enhancement. A

comparative analysis is required in order to improve the

visualization range of OLAP data.

Followed by the above advanced visualization technique,

Sifer [10] introduced a visual interface technique for

exploring OLAP data with coordinated-dimension

hierarchies. They identified the problem that a number of

text based and visual interfaces for querying multi-

dimensional data exist but many of these interfaces

are not applicable to OLAP. The reason is the lack of

support for the use of dimensional hierarchies for

selection and aggregation. The main contribution is the

progressive view coordination interface which provides

better support for query refinement than existing

interfaces, by helping users decide the next query step

with intermediate result overviews, and also helping the

users to change a previous selection decision with

retained selection context views. The strength of their

work is the implementation of SGViewer visualization

tool for their technique using a web log dataset. The

limitation is that a very simple and small (only 3

dimensions) web log dataset was used for the illustration

and, to use this technique the data has to be converted

into Structured Graph Format Extensible Mark-up

Language (XML) document. Therefore, it is a specific

data format conversion limitation.

Voss et al. [11], adopted another way and suggested an

extension in an advanced tool (CommonGIS) for the

highly interactive visual exploration of spatial data. The

identified problem was that the CommonGIS tool

lacks the connection to OLAP warehouse to be a

complete Business Intelligence (BI) application. They

explored how to connect the tool to OLAP warehouses as

another source of multi-dimensional data and designed

architecture for the extension of CommonGIS. The

strength of their work is that they allowed support of

MDX (Multi-Dimensional Expression) for the

manipulation of cube data using the application builder of

the tool. The limitation of their work is the manual script

writing process to convert client data into a compatible

format so that the data is incrementally transmitted from

the client source to the OLAP data cube.

In the same visualization enhancement area, Scotch et al.

[12] proposed a tool for Spatial OLAP, called SOVAT

(Spatial OLAP Visualization and Analysis Tool). The

authors identified the problem that for community health

research a comprehensive and thorough analysis is

needed for effective public health evaluation. The

major contribution of their work is the development of

tool for community health assessments. The strength of

the tool is that it combines all necessary features of

OLAP and GIS (Geospatial Information Systems) into

one system to support comprehensive community

health decisions. The limitation of this system is that it

allows the visualization of spatial data and only permits

the analysis on numeric data; however while performing

analysis on spatial databases the researchers require tool

support for not only number data but also other complex

data types such as images, videos and maps. The

community health researchers are more interested in

the patterns or areas of an image for the evaluation

purposes, rather than exploring numerical analysis of the

data using OLAP.

Cuzzocrea et al. [13] introduced a hierarchy-driven

compression technique for the advanced visualization of

multi-dimensional cubes. This technique relied on the

facility of generating a “semantic-aware” compressed

representation of two-dimensional OLAP views. The

strength of this particular technique is that it is most

suitable for the handheld-devices where compression in

the data cubes is required. The limitation is that it only

supported two-dimensional OLAP views but, in real

scenarios, there are multiple dimensions and facts in the

dataset. The technique is more suitable for MOLAP

scenarios and not very effective in the case of ROLAP,

where multiple attributes exist in the relational dataset

and the hierarchies are more complex within each

dimension.

In 2005, [14] presented a new visual interactive

exploration technique for OLAP. This work is similar to

this paper work in terms of OLAP user facilitation. This

technique and the proposed enhanced architecture, allows

users who have less knowledge of OLAP technology to

explore and analyze OLAP data cubes without generating

sophisticated queries. The following year, [15] proposed

a framework for querying complex multi-dimensional

data with a major effort applied to transforming irregular

hierarchies to make them navigable in a uniform manner.

Lastly, in 2008, [16] introduced a comprehensive visual

exploration framework which implements OLAP

operations in form of powerful data navigation. Users can

explore data using a variety of interactive visualization

techniques using the proposed framework.

Research communities have often neglected the data-

visualization problem in the past [13]. But we note that a

considerable amount of work is recently being performed

by various research communities on the visualization of

OLAP data which indeed can play a very vital role in

effective analysis of data. Our work in this paper is

similar to the above mentioned contributions as it also

improves the performance of OLAP in terms of cube

construction time. Furthermore, we are using cube

visualization API, which dramatically enhances the visual

JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011 77

interactive exploration of cube data with the help of

charts, graphs and data grids at the front end. In the next

section, we present some noteworthy work done on the

integration of OLAP enhancement techniques.

B. Integrated enhancement of OLAP

In this section, we highlight the research done on the

integrated enhancement of OLAP performance and its

data visualization. Although some of the previously

discussed research work proved the enhancement of

OLAP systems but none of the work was intended

towards focusing on the integrated enhancement of

OLAP performance and visualization and there exists a

strong need for it [17]. Asghar et al. [7] proposed a

functionality-enhancement technique using Self-

Organizing Neural Networks. To some extent this

technique proposed the integrated enhancement of OLAP

by passing the mined data to the OLAP engine for a

more focused analysis and hence adding intelligence

to the OLAP system. The major limitation of the work

was the deficiency of enhancement on the visualization

aspects of OLAP systems. Their enhancement

architecture doesn’t provide satisfactory answers for the

visualization of complex data sets. During the last few

years, a lot of work has been done on the enhancement of

OLAP particularly on its performance and visualization.

Various architectures [7, 68, and 69] were proposed for

the one dimensional enhancement of OLAP systems. But,

the most relevant work in this integrated enhancement

was done by [70]. The authors proposed architecture for

the integrated enhancement of OLAP system. They have

used visualization tools and developed a prototype which

improved performance in terms of cube construction

time. Both performance and visualization aspects were

catered by the proposed architecture. The limitation of the

work was again the dependency of GSOM to produce

hierarchical clusters and the manual schema generation of

the hierarchical cluster tables. The proposed architecture

has a gap which is the transformation of cluster

relationship table to OLAP schema. Authors only

validated the prototype using only one type of OLAP

schema that is the star schema. The paper work is

different from the above mentioned literature as we are

not using GSOM for hierarchical clustering. Furthermore,

a schema builder component has been proposed which

takes clustered data and generates three types of OLAP

schemas automatically. Additionally, cube has been

constructed which holds the clustered data to form an

OLAM cube to support the process of online analytical

mining which was missing. As this paper deals with the

integration of OLAP and data mining, in the next section

some important work on the combination of OLAP and

data mining has been reviewed.

C. Combining OLAP with Data Mining

Several previous studies [5, 7, 8 and 23] emphasized the

likelihood and interest of coupling OLAP and Data

Mining. Efficient and effective data mining integrates the

concept of [24], which is the integration of OLAP and

OLAM systems. OLAM is a promising direction for

mining knowledge from multidimensional databases [23].

Similarly, Hua [25] proposed and developed an

interesting association rule mining approach called

Online Analytical Mining of association rules. It

integrated OLAP technology with association rule mining

methods and leads to flexible multidimensional and

multi-level association rule mining. Similarly, Josph et al.

[26] presented a methodology that derives the association

rule of web pages tick sequences according to the support

level and confidence level of user requirements. This

methodology was responsible for identifying a set of

frequently accessed web pages on a website by a user.

The result is the list of potential customers for a certain

product or service on a target web page.

Dzeroski et al. [27] combined OLAP and Data Mining in

a different way to discover patterns in a database of

patients. Two data mining techniques, clustering and

decision tree induction were used. Clustering was used to

group patients according to the overall presence/absence

of deletions at the tested markers. Decision trees and

OLAP were used to inspect the resulting clustering and to

look for correlations between deletion patterns,

populations and the clinical picture of infertility.

Dehne et al. [28] studied the applicability of coarse

grained parallel computing model (CGM) to OLAP for

data mining. Authors presented a general framework for

the CGM which allows for the efficient parallelization of

the existing data cube construction algorithm for OLAP.

Experimental data showed that this approach yield

optimal speed up even when run on a simple processor

cluster via a standard switch. The study shows that OLAP

and data mining, if combined together, can produce

greater benefits in a number of diverse research areas.

Our proposed work is similar in terms of combining

OLAP and data mining techniques but our main focus is

on a particular data mining technique known as

Hierarchical Clustering. Furthermore, we are using data

mining as a pre-processing step to get better

understanding of data before passing it to the automatic

schema builder and which then generates schema for

OLAP engine. In the next section, the related data mining

technique which is hierarchal clustering and its usage

with OLAP is the centre of attention.

D. Use of Hierarchical Clustering with OLAP

A number of data mining techniques which can be

combined with the OLAP technology for the better

solutions of research problems were described in the

previous section. Hierarchical clustering of data is a well

know mining technique and it has been used by a number

of research communities in the past. In this paper

hierarchal clustering has been used as a data mining

technique with OLAP. In this section a review of

hierarchical clustering in conjunction with OLAP

technology has been performed.

78 JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011

Markl et al. [29] suggested that OLAP performance can

be improved by using the Multidimensional Hierarchical

Clustering (MHC) technique. Clustering was introduced

as a way to speed up query aggregation without

additional storage cost for view materialization. The

authors identified the problem with queries which either

select a very small set of data or perform

aggregations on a fairly large data set. The sole

contribution of their work is an encoding scheme for

hierarchical dimensions that enables clustering of data

with respect to multiple, hierarchical dimensions. The

major strength of the work lies in the comparison of

their MHC technique with the traditional bitmap

indexing approach on the real world data (7GB in

size) and finding an increase in the performance up to

the factor of 10.

Markl et al. [30], again utilized hierarchical clustering in

the context of OLAP queries. Authors investigated the

impacts on query processing in RDBMS when using UB-

Trees and multidimensional hierarchical clustering for

physical data organization. Furthermore, they illustrated

the benefits achieved by performance measurements of

queries using star schema for a real world application of a

SAP business information warehouse.

Karaynnidis et al. [31], proposed a novel multi-

dimensional file structure for organizing the most detailed

data of the cube, the CUBE file. The CUBE file archives

hierarchical clustering of data enabling fast access via

hierarchical restrictions. It imposes low storage cost and

adapts perfectly to extensive sparseness of data space.

Results show that the CUBE File outperforms the most

effective method proposed up to now for hierarchically

clustering the cube, resulting in 7-9 times less I/Os on

average for all workloads tested.

In the same context, Theodoratos et al. [32] claimed the

Heuristic Optimization of OLAP in MHC (Multi-

dimensionally Hierarchically Clustered) databases. They

identified the problem that commercial relational

database management systems use multiple one-

dimensional indexes to process OLAP queries that

restrict multiple dimensions. They presented architecture

for MHC databases based on the star schema called

‘CSB’ star. The focus was to facilitate the user to query

the typical star schema and to formulate the given query

for ‘CSB’ star schema using query processor. The

limitation of their work was that they only tested a

particular class of typical OLAP queries over the CSB

star schema to check efficiency.

Tsois et al. [33], proposed cost based optimization of

aggregation star queries on hierarchically clustered data

warehouses. Hierarchically clustered data warehouses

were used to apply a cost based method for the optimal

application of the pre-grouping transformation. Authors

identified the most suitable algorithms for the operations

related to pre-grouping and derived detailed cost formulas

for them.

It is apparent from the review that various authors have

used hierarchical clustering technique in different ways

with OLAP to improve the functionality and to get better

analytical results. The major difference of this paper work

with the reviewed work of this section is that our use of

hierarchical clustering is to generate hierarchal clusters to

get an insight of the data before passing it further for

OLAP operations. This mining technique has not been

applied on the cube for archiving cube data or for the

optimization of OLAP queries. Instead of query

optimization, OLAP performance has been improved in

terms of cube construction time through the use of

hierarchically clustered data.

E. Automatic OLAP schema generation

In this section, we discuss the major research work on the

automation process for the generation of OLAP schema.

Hann et al. [34], proposed the generation of tool specific

schemata for OLAP from conceptual graphical models.

Their work described the design and implementation of

the generation component in the context of their own

Bablefish data warehouse environment. The principle

issues of designing and implementing such a automatic

schema generation component and the possible solutions

have been discussed by the authors. Further topics are the

use of graph grammars for specifying and parsing

graphical multi-dimensional schema descriptions and the

integration of the generation process into a metadata

cantered modeling tool environment.

Peralta et al. [35], highlighted the existing work in the

area of automation and claimed that automation is

focused towards data models, data structures specifically

designed for Data warehouse (DW), and criteria for

defining table partitions and indexes. The major research

contribution of the authors is a step forward towards the

automation of DW relational design through a rule-based

mechanism, which automatically generates the DW

schema by applying existing DW design knowledge. The

proposed rules embed design strategies which are

triggered by conditions on requirements and source

databases, and perform the schema generation through

the application of predefined DW design oriented

transformations.

For the sake of accomplishing the ease in the automation

work, Tryfona et al. [36], built a conceptual model

(StarER) for data warehouse design on the basis of user

modeling requirements. The StarER model combines the

star structure, which is dominant in data warehouses, with

the semantically rich constructs of the ER model.

Comparison of the proposed model with other existing

models has been performed, pointing out differences and

similarities. Examples from a mortgage data warehouse

environment, in which starER is tested, that revealed the

ease of understanding of the model, as well as the

efficiency in representing complex information at the

semantic level.

JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011 79

For the purpose of achieving the automatic generation,

Song et al. [37] introduced an automatic tool for

generating star schema entity-relationship diagram

(ERD). A prototype named SAMSTAR was presented,

which was used for the automatic generation of star

schema from an ERD. The system takes an ERD drawn

by ERwin Data Modeler as an input and generates star

schemas. SAMSTAR displays the resulting star schemas

on a computer screen graphically. With this automatic

generation of star schema, this system helps designers

reduce their efforts and time in building data warehouse

schemas.

It is evident from the literature on automatic generation of

schema that none of the previous works used

hierarchically clustering to generate OLAP schema.

Previous work on automation has a number of limitations

as well such as support for only one schema type,

dependency of ER diagram modeling and having to

design rule based mechanism before the automation

process.

This work is different as compared to existing automation

work in the literature. Firstly, the results of hierarchical

clustering are used to generate any of the three types of

schemas. Furthermore, the work is not limited to

designing the schema but it also populates the data from

the clustered data source to the automatically generated

schema tables. In addition to this, the schema builder also

communicates with database server and deploys the

generated schema in the server as a complete database. In

section 3.6, comparison has been made of this paper work

with the four themes of the literature review. The purpose

of this comparison is to emphasize the significance of our

work and to highlight the existing limitations in the

related work.

F. Comparison of the proposed work with the

literature reviewed

In the previous sections the individual papers were

discussed. Now, this section presents a comparison of this

paper work with the four major themes of the literature

reviewed. Firstly, a substantial amount of work has been

done on the enhancement of OLAP in terms of its

performance and visualization. According to [5, 6 and 7],

performance of OLAP systems can be improved with the

help of different techniques like parallel computing,

indexing and materialization, neural networking and

agglomerative hierarchical clustering. This paper work is

a bit similar in terms of OLAP performance improvement

but we have a broader scope as in this paper, we are

proposing an architecture which includes the OLAP’s

data visualization enhancement as well.

In the area of OLAP’s visualization, research

communities has also paid attention and reported the

work on visualization enhancement in [9, 10, 11, 12, 13,

14, 15 and 16]. Visualization enhancement techniques

like Cube Presentation Model (CPM), visual interfacing,

interactive exploration, hierarchy driven compression and

complex multi-dimensional querying were introduced.

Previously it has been identified that there is a strong

need of integrating performance improvement and

visualization enhancement techniques [17]. Furthermore

by utilizing the work of [7], which lacked in the areas of

visualization work of OLAP data, schema generation and

cube processing, this paper work moved a step forward

and introduced an advanced architecture for the

integrated enhancement of OLAP by combining OLAP

with data mining techniques in [70].

Various researchers have contributed in the combination

of OLAP and data mining in the past. Significant works

in this area include [24, 25, 26, 27 and 28]. These authors

used different data mining techniques in their areas of

interest and produced better results by combining OLAP

with data mining. The two major techniques of data

mining used in this regard were Association Rule Mining

and Hierarchical Clustering. In this paper we have used

the hierarchical clustering technique for the integrated

enhancement of OLAP’s performance and visualization.

Hierarchical clustering has been used by [9, 30, 31, 32

and 33] in conjunction with OLAP systems for a number

of reasons. Some authors used this technique for query

optimization, physical data organization, fast data access

and reduced storage cost. The work in this paper is

different from these as we used hierarchical clustering as

a pre-processing step of data mining. Furthermore, we are

utilizing it to produce hierarchical relationship data for

the automatic generation of OLAP schema.

As far as the automatic generation of OLAP schema is

concerned, [34, 35, 36 and 37] reported their work in this

area. To our knowledge none of these works used

hierarchically clustered data to generate OLAP schema.

Most of the work is using ER diagrams or conceptual

graphical models to produce only one type of schema

which is mostly star schema. Another major limitation in

the work is that the tools and techniques used for

automatic generation of schema only draws of identify

the structure of schema and doesn’t populate the

corresponding data in the generated schema.

This paper work is not only generating OLAP schema of

three types namely; star, snowflake and galaxy but our

automated schema builder tool provides automation in

uploading the data from clustered file to relevant portions

of the schema. This feature along with other

specifications distinguishes the work done in this paper

with others. But in a nutshell the paper work is intended

to support the OLAM in a distinctive way. It is distinctive

because none of the previous work dealt with all the four

major aspects of our proposed research.

IV. PROPOSED ARCHITECTURE

The architecture for integrated OLAM has been proposed

in this section. But before moving forward towards the

architecture itself, it is important to emphasize the

necessity of the proposed architecture that why exactly do

we need to have architecture for the integrated OLAM?

To answer this important question we begin the

explanation starting with the overview of the previous

80 JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011

Figure 1. Proposed architecture for integrated OLAM

section. As discussed earlier, it is evident from the

literature review that none of the proposed work in the

past had targeted towards the integration of enhanced

OLAP with data mining along with the automation in the

schema generation process. Integrated OLAM is the

combination of data mining and enhanced OLAP with the

incorporation of automatic schema generation from

mined results. A number of authors [5, 7, 8, 23, 24 and

25], in the recent years, have emphasized on the concept

of combination of OLAP and data mining. Although this

previously reported work justified the advantages of such

combination but the major limitation in their work is the

absence of enhanced OLAP system combined with data

mining. Typically, a standard OLAP system is integrated

with data mining. In parallel to this, various authors [6, 9,

10, 11, 12, 13 and 14] reported their work on the

enhancement of OLAP in terms of its performance or

data visualization. In this paper, it is recommended that

such an enhanced OLAP system should be integrated

with data mining so that the benefits of enhancement

achieved in OLAP are combined with data mining

techniques to produce overall improvement in the OLAM

system. In addition to this distinctive combination, the

integrated OLAM architecture has been extended with the

additional layer for automatic schema generation. For the

purpose of this transformation we suggest an independent

automatic schema generation module that fulfils both the

requirements of schema generation and transformation of

data. Based on the above discussion, there is a strong

requirement of a significant architecture that takes the

data mining results to automatically generate schema,

transform mined data in the schema, incorporate

enhanced OLAP in it, support OLAM along with

overcoming the existing limitations and lastly, produce

improved, focused and intelligent results. Figure 1

illustrates the proposed architecture for integrated

OLAM.

The goal of the proposed architecture is to integrate

enhanced OLAP with data mining and to provide

automatic schema generation from the mined data. We

elaborate the four distinct layers of the proposed

architecture which are 1) Data Mining: pre-processing

layer, 2) Automatic schema generation layer, 3)

Enhanced OLAP layer and 4) User interface or Front-end

layer. Each layer along with its components is discussed

in separate sub sections. To begin with the explanation,

we start with the data mining layer.

JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011 81

A. Data mining: Pre-Processing Layer

This layer is responsible for the data mining process. The

three components in this layer are as follows:

• Dataset: This component represents the data set.

A large number of data set exist and this

component is responsible for the selection of a

particular data set on which data mining is to be

performed.

• Hierarchical Clustering: Once the data set is

selected, the data mining technique which is

hierarchical clustering is applied in this

component on the data. This component also

handles clustering parameters which include

linkage method (Single, Complete or Average),

clustering directions (rows, columns, both) and

similarity/distance measure (Euclidean,

Manhattan, Simple matching for nominal or

Pearson correlation coefficient). On the given set

of clustering parameters this component produce

the hierarchical clusters.

• Clustered Data: The hierarchical clusters

generated by the previous component are stored

and saved in the form of clustered data. This

component of the layer also saves the hierarchy

relationship among the data along with the

actual data in each cluster. The hierarchically

clustered data become the input of the next layer

which is responsible for the automation process

of schema generation.

B. Automatic schema generation layer

The purpose of this layer is to automate the process of

schema generation. To achieve this purpose the data

mining layer feeds the mined data or clustered data to this

layer for the generation of schema from clustered data

set. The layer has distinct components and the working of

each component is explained as follows:

• Schema Builder: It is the major component of

schema generation layer and the sole function of

this component is to build a particular type of

schema for the clustered data set. As this

component has to perform numerous schema

generation activities, therefore it has a number

of sub-components to achieve the purpose of

building the schema. The sub-components of the

schema builder are as follows:

o Clustered Data Reader: This sub-

component of the schema builder reads

the data generated by the hierarchal

clustering and save it temporarily for

future operations.

o Clustered Data Viewer: It allows

viewing the clustered data read by the

reader in the grid view. All the cluster

hierarchy and the number of records

present in each can be viewed through

it.

o Schema Generator: After the

clustered data has been read and viewed

the data goes to the component called

schema generator. The sub-components

of the schema generator are explained

below.

 Dimensions and Facts Identifier:

This is the first sub-component of

the schema generator. In the

generation process of schema the

first important step is the

identification of dimensions and

facts from the clustered data. For

this important purpose the data in

the schema generator goes to the

Dimension and facts identifier,

which identify the dimensions and

facts present in the clustered data.

It also identifies the cluster

hierarchy and the numeric type of

data which are the facts or key

performance indicators of the data.

 Dimensions and Facts Separator:

The identifier component just

identifies the various dimension

and facts present in the clustered

data. But in real the dimension and

the facts are to be separated in

order to build the schema. For this

purpose the Dimension and facts

separator splits the identified

dimensions and facts to be used for

different types of schema

generation.

 Schema Type Selector: At the

stage when dimensions and facts

are identified and separated the

Schema type selector component

handles the designing of a specific

type of schema. This selector

allows the selection of a particular

type of schema which can be a star,

snowflake or galaxy.

 Warehouse Builder: After

determining the schema type to be

generated the dimensions and facts

along with the specific schema

type goes in to another major and

core unit of schema builder which

is called the warehouse builder.

The warehouse builder as its name

implies, builds a warehouse of a

particular schema type and consists

of the following main sub-

82 JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011

components. The functionality of

each of the component in the

warehouse builder is described

below:

• Dimensional hierarchy

handler: It is the first

component of the

warehouse builder and it

serves the purpose of

handling the hierarchy in

the dimensions. As the

clustered data has its own

hierarchy and the

dimensional data

identified and separated

previously have to be in

some hierarchical order

within each dimension.

Because the dimensional

hierarchy is different for

different type of schema

that is why this handler

component handles the

hierarchy present within

each dimension.

• Dimension and fact table

creator: Because the

warehouse builder

automatically creates a

warehouse in the form of

a database so the creation

of the dimension and fact

table is a must. For this

reason the dimension and

fact table creator takes

dimensional hierarchy

from the previous

component and the facts

separated by the separator

component and create the

tables according to the

specific schema type.

Within each table created

for a dimension the

hierarchy structure is

determined and created

using the hierarchy

handler output, which is

different for each type of

schema.

• Table relationship

manager: For each type

of schema the referencing

of dimension and fact

table is different. For star

the dimensions are

connected directly, in

Snowflake schema there

exists a normalized

relationship between the

dimensions whereas for

Galaxy both types of

relationships are being

made. To handle this

complexity of relationship

among dimension and fact

tables created by the

warehouse builder for a

specific schema type, the

table relationship manager

plays its part and manages

all kind of relationships

among the tables

accordingly.

• Data population unit:

When all the tables are

created along with the

hierarchy within each

dimension and the

relationship is being

created among the

dimensions and fact

tables, the data population

unit comes in to play. The

role of this unit is to load

the clustered data in the

schema intelligently.

Therefore, this unit picks

the clustered data and

inserts it in specific

columns of the dimension

and fact tables so that this

data can be used later for

the OLAP operations that

are to be performed in the

next layer of the

architecture.

• Schema Viewer: This

component takes the

generated schema from

the schema generator and

gives the view of the

automatically generated

schema in the form of

schema diagram.

C. Enhanced OLAP Layer

As our proposed work is the integration of data mining

with enhanced OLAP, so in this section we discuss the

layer which is exclusively responsible for the OLAP

enhancement in the proposed architecture. Furthermore,

the construction of cube from clustered data in the form

of automatically generated schema is presented. The

various components of this layer along with their internal

functionality are as follows:

JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011 83

• Database Server: The main purpose of the

database server is to store the automatically

generated databases from the warehouse builder

of the schema generator. The components of a

database server which are used for the

communication of architectural layers and

internal function are explained below:

o Generated Databases: This

component stores the automatically

generated schemas in the form of

databases in the database server. For

each kind of schema a specific database

is automatically created along with all

the dimensions and fact tables.

o External Component API: The

external component API is responsible

for the communication involved in the

creation of specific type of database

within the database server. It uses the

requests from the schema builder

component and responds accordingly.

o Storage Engine: This component of

the database server handles the storage

allocation of the generated databases in

the database server. It allocates

memory type for the created database

and handles the size of the particular

database in the memory.

o Query Processor: warehouse builder

of the schema generator component

discussed in the previous layer send

requests using the connection string to

the database server in the form of

queries. The query processor

component reads the request sent by the

warehouse builder and processes the

queries. These queries when processed

create the databases and the tables with

specific columns along with the

clustered data.

• OLAP Server: Similar to the database server

which stores and provides a database

management system OLAP server is another

important component of this layer that handles

the cube construction, storage and manipulation.

OLAP server’s core purpose is to handle the

cube related queries and manipulations. The

component worth discussing in this are as

follows:

o Data Source: This component of the

OLAP server manages various sources

of data which are to be used for cube

construction process. In the proposed

architecture of integrated OLAM,

automatically generated schema

databases become the source in the

OLAP server for cube construction. It

means that the generated databases are

connected with data source using this

component for future use. So the

storage of data is in the database server

and the connection between the two

servers handles the communication

involved form one server to another.

o OLAP Engine: This component allows

the construction of cube on a specific

data connected with the data source

component. It is the most powerful

component of the server which handles

the processing of cubes which involve

aggregation of data at various

hierarchical levels.

o Cube Storage: Once data source

makes a connection with the data for

the cube construction and the OLAP

engine construct the cube. The cube is

stored in the OLAP server and the

component responsible for this is the

cube storage component. It also allows

the type of cube storage which can be

MOLAP, ROLAP or HOLAP.

• Cube Visualization API: After the construction

and processing of OLAM cube the cube

visualization API comes in to play. The purpose

of this component is to apply the cube data

visualization methods to enhance the power of

OLAP visualization. This component of the

layer provides rich interactive analysis functions

and operations that enhance the capabilities of

standard OLAP system.

D. User Interface Layer

This layer represents the user interface of the proposed

architecture. At this layer, the OLAM cube, which is

constructed and processed in the previous layer, allows

targeted and focused analysis. The constructed cube is

named as OLAM cube because it consist of mined data

results of the hierarchical clusters with the aggregation of

facts present in each cluster. It allows all the OLAP

operation to be performed on it such as drill-down, rollup

etc. The OLAM cube data goes through the visualization

API which enhances the visual exploration capabilities of

its data. It allows interactive and visual analysis on the

OLAM cube data. Therefore, the user gets a real insight

of this mined data at this user-interface layer.

The user interface layer represents the targeted and

focused analysis. It is targeted and focused analysis

because at the first layer of data mining when the clusters

are generated through hierarchical clustering, user had a

view to determine the cluster of his/her interest and

84 JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011

continue to the second layer of automatic schema

generation for that specific cluster. Moving on with the

architectural flow, the schema of user’s choice is

generated automatically and cube is constructed only for

the schema of that particular cluster. This is the reason

why the final cube itself has the targeted data, therefore

the analysis or visual exploration performed at this stage

is on a focused and targeted data and not on the complete

data set.

E. Benefits Achieved

There are a number of significant benefits achieved

through the proposed architecture. The list of these

benefits are as follows:

o Use of data mining as a pre-processing step

for the targeted analysis of data.

o Allowance of various hierarchical clustering

algorithms and data linkage methods.

o Automatic generation of three types of

schema from the clustered data.

o Complete diagrammatical view of the

generated schema using schema viewer.

o Use of distributed database and OLAP

servers.

o Dynamic creation of separate databases for

each type of generated schema.

o Performance improvement of cube in terms

of its construction time.

o Visualization enhancement for the

interactive and powerful visual exploration

of cube data through the use of visualization

API.

o Construction of OLAM cube instead of

simple OLAP cube.

o Powerful front-end for the focused and

interactive analysis on the OLAM cube.

V. ARCHITECTURAL IMPLEMENTATION

In this section, based on our implementation, we discuss

in detail the steps involved in the implementation of the

proposed architecture. Some tools and technologies have

been used for the implementation of each layer of the

proposed architecture and divided the implementation

phase into five major steps. Figure 2 depicts the major

steps of architectural implementation.

Each step of implementation phase is discussed

individually in the following sub sections.

A. Hierarchical Clustering of Data

The proposed architecture starts with the layer of data

mining and the technique used for data mining is the

hierarchical clustering. The first step of the

implementation is the hierarchical clustering of data. To

serve the purpose of step 1, we used Hierarchal

Clustering Explorer (HCE) tool for generating the

hierarchical clusters of data. This tool takes input data file

and allows the hierarchical clustering of given data based

on different clustering parameters. Figure 3 shows the

interface of cluster parameter selection form, which has

the different parameters for selecting the type of

hierarchical clustering.

At this point, user can select the parameters of his/her

choice to perform specific type of hierarchical clustering

on the data. Upon the selection of clustering parameters,

the data as a result, is hierarchically clustered and the

hierarchy details are shown in the HCE tool.

Furthermore, the tool allows exporting the clustered data

results in the form of text files. At this stage, after the

clustering has been performed the results are exported

which serve as the input of the next step of automatic

schema generation. We elaborate the next step of

automatic schema generation in the following section.

Figure 2. Sequence of architectural implementation steps

JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011 85

Figure 3. Clustering parameter selection in HCE tool.

B. Automatic Schema Generation

In this step, the cluster results file which was exported

using the HCE tool becomes the input of the automatic

schema builder. We developed a prototype to automate

the process of schema generation. The prototype takes the

clustered data and generates the schema of any particular

type such as star, snowflake or galaxy/constellation. The

prototype has been developed using the C sharp (C#)

programming language in Microsoft Visual Studio.net

2005. The implementation logic used for the development

of automatic schema builder is explained in the form of a

main algorithm 5.2.1.

Algorithm 5.2.1

iinto j

(__)

(, )

//* 1:Re _ _

_( )

R )

(

r r

H C D hie ra rchic a lly C lustered D a ta

Schema star snowflake or galaxy

Step ading H ierarchically Clustered data HC D

nput

Output

M ethod

while

ead

≠∅

w h ere len is th e len gt

let be colum n headear

if (i ) then

fo r (j= 1 to r.len ) h

∈∅

Dm j

else

fo r (j= 1 to r.len )

kHC

←

86 JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011

( )

if then

/*len is the lenght

for (j= 1 to C .len)

//* 2: _ _ _ )

(

while

Step Viewing HCD in dataviewer DV

r≠∅

∋

of the string

)

send C j to D

else

for (k=1 to C D .len

Dk V

D be the dimension storage variable and let F be the fact variable

send C to D

while (r )

//* 3: & _ _ _

let

Step Dimension fact identificatin and Seperation

≠∅

i+1

I) /*w here I is an interger

call pro cedue D im _sep( i, i ) /*w here (ri and ri+1) D

if (r

∈

else

call F act_sep(ri) w here (ri) F

i i +2

procedure Dim_sep(var1,var2) /* var1 and var2 temporary va

∈

←

riables

[ , ] [ , 1]

Fact_sep(var3)

F[p]

//* 4: _ _se

D m n D ri ri

procedure

Fri

let

Step Schema type lection

←+

←′[]

be the selected schema type

if (S "star")

call procedure (Genereate_star)

else

if (S "snow flake")

←

call procedure (Generarate _snow flake)

else

if (S "galaxy")

call procedure (G enerate_galaxy)

←

en d if

DT be the dim ension tables and FT be the fact table

if (S T " s tar")

or i

//* 5: _ _

let let

Step V iew in g gen erated schem a

←

f (S T "sn ow flak e ")

dim en sio n ta b le s in to D T

w h ile ( D T i )

se n d

read

←

≠∅

D T i to T V */ SV is the table view er

rea d fac t ta b le s in to F T

se n d F T to T V

else

if (S T " ")

dim en sio n s in to D T

en d if

galaxy

read

←

( )

D T i fo r ta b le v iew

rea d fa c t ta b le s in F T

w h ile D T i

send

≠∅

w h ile (F T i )

se n d F T i fo r ta b le v iew

en d all.

≠∅

JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011 87

In the above algorithm the basic steps of automatic

schema generator are represented using an algorithm.

C. Database and OLAP server deployment

In the previous step, automatic schema has been

generated and the schema builder actually creates

database automatically in the database server to hold the

schema along with dimension and facts tables. To store

the automatically generated database by the schema

builder in the server, we deployed Microsoft SQL server

2000.This database becomes the source data for the cube

creation process. The cube is constructed in the OLAP

server. We deployed Microsoft analysis service 2000 for

the purpose of OLAP operations from the clustered

database.

D. OLAM cube construction

The fourth step of the implementation is the construction

of OLAM cube. The database server holds the

automatically generated schema of clustered data. The

OLAP server connects with the database server to get

clustered data from the database. Cube wizard (of the

Microsoft analysis services 2000 was used to construct

the OLAM cube. It is an OLAM cube because it has the

mined data aggregates in it. User can perform the

standard OLAP operations on this OLAM cube.

E. Embedding Cube Visualization API

In the previous section, MS Analysis services 2000 was

introduced to serve the purpose of OLAP server and to

construct OLAM cube using its cube wizard option. The

emphasis in this paper from the beginning is to integrate

enhanced OLAP system with data mining technique and

automation in schema generation. After the elaboration of

implementation of mining technique and the automation

process of schema generation but at this point, we

embedded the cube visualization API called Dundas

OLAP services [71] which provides user-friendly and

powerful visual data exploration controls.

Using the above mentioned controls, we developed

another prototype to view the constructed OLAM cube.

Figure 4 shows the user interface of our prototype which

has all the above mentioned controls to work on cube

data. It can be seen from the Figure 4 that the controls

allow users to perform rich analysis tasks on the data.

Furthermore, with the help of these controls, interactive

visual exploration of data in the form of drill-down and

roll up charts , grids and selection menus is possible.

Figure 4. User interface of prototype with all the visualization controls

88 JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011

VI. VALIDATION AND EVALUATION

n this section, we validated and evaluated the proposed

architecture with the help of experiments. We used a real

life data set (Census [73]) to perform experiments and

discussed the results of these experiments. A comparative

analysis of the proposed architecture with previous work

is discussed at the end of this section. Such analysis

highlights the benefits gained through the proposed

integrated Online Analytical Mining (OLAM)

architecture. The experimental details are presented in the

following sub section.

A. Experiment

In this experiment, we use US Census dataset [73], The

data was collected as part of the 1990 census. There are

68 categorical attributes. This data set was derived from

the USCensus1990raw data set [74]. The attributes list

and the coding for the values is described below. Many of

the less useful attributes in the original data set have been

dropped, the few continuous variables have been

discretized and the few discrete variables that have a

large number of possible values have been collapsed to

have fewer possible values. More specifically the

USCensus1990 data set was obtained from the

USCensus1990raw data set by the following sequence of

operations;

- Randomization: The order of the cases in the original

USCensus1990raw data set was randomly permuted.

Selection of attributes: The 68 attributes included in the

data set are given below. In the USCensus1990 data set

we have added a single letter prefix to the original name.

We add the letter 'i' to indicate that the original attribute

values are used and ’d’ to indicate that original attribute

values for each case have been mapped to new values.

Hierarchies of values are provided in the file

USCensus1990raw.coding.htm [75] and the mapping

functions used to transform the USCensus1990raw to the

USCensus1990 data sets are giving in the file

USCensus1990.mapping.sql [76]. The data is contained

in a file called USCensus1990.data.txt [77].

In order to do the first step of this experiment, we passed

this data set to the HCE tool so that we can get the

hierarchical clusters from this huge data set. Similar to

the previous experiment we set the clustering parameter

for the hierarchical cluster generation process and at the

top level we got five distinct clusters having individual

hierarchies as depicted in Figure 5.

It can be seen from the Figure 5 that there are five

clusters in the dendogram view of HCE tool. At the

bottom of the hierarchical representation we can also see

the data present in each cluster. The actual data can be

extracted for each cluster as well. The next step of the

experiment is to transform this hierarchically clustered

data which will be fed into the developed prototype called

the Automatic Schema Builder.

At this stage of the experiment, individual clusters data is

being analyzed and each of the cluster is given a name

which classify its characteristics. The data set has records

of the persons, for that reason we classified the clusters

on the basis of person’s industry and occupation. The

hierarchical clusters were named accordingly.

Figure 5. Hierarchical clusters of census dataset

JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011 89

The naming convention used is similar to the previous

experiment such as the first letter of the industry or

occupation has been used as an abbreviation. These

abbreviations are represented in the Table 1.

This hierarchical information along with the clustered

data has been fed into the automatic schema builder. The

same procedure follows in the schema builder; it displays

the cluster data and further asks for the type of schema to

be generated. In contrast with the previous experiment we

chose snowflake schema for generation. The automatic

schema builder ran the same sequence of steps for

snowflake schema generation explained in the snowflake

schema generation algorithm.

TABLE 1 ABBREVIATION OF CENSUS DATA CLUSTERS

Cluster Names Clus_abb Cluster Names Clus_abb

Electric,Electronic and Computing (EEC) C1 Businessmen (BM) C3-1

Electric,Electronic and Computing (EEC) C1 Labours and Workers (LW) C3-2

Electric,Electronic and Computing (EEC) C1 Copiers and Engineers (CE) C4-1

Electric,Electronic and Computing (EEC) C1 Publishers and Designers (PD) C4-2

Agriculture,Foresty and Fishery (AFF) C2 Publishers and Designers (PD) C4-3

Agriculture,Foresty and Fishery (AFF) C2 Constructors and Architects (CA) C5-1

Rubber, Leather and Furniture (RLF) C3 E.Enginers C1-1-1

Rubber, Leather and Furniture (RLF) C3 Elec.Enginers C1-2-1

Rubber, Leather and Furniture (RLF) C3 Sr.Teachers C1-3-1

Rubber, Leather and Furniture (RLF) C3 Jr.Teachers C1-3-2

Printing, Publishing and Textile (PPT) C4 Retailers C1-4-1

Printing, Publishing and Textile (PPT) C4 Salesmen C1-4-2

Printing, Publishing and Textile (PPT) C4 Graduates C1-5-1

Printing, Publishing and Textile (PPT) C4 Scientists C2-1-1

Machinery and Construction (MC) C5 Policy makers C2-1-2

Machinery and Construction (MC) C5 Analysts C2-2-1

Engineers and Specialists (ES) C1-1 Owners C3-1-1

Administrators and Managers (AM) C1-2 Carpentars C3-2-1

Teachers and Trainers (TT) C1-3 Swievers C3-2-2

Teachers and Trainers (TT) C1-3 Taylors C3-2-3

Retailers and Salesmen (RS) C1-4 Copier C4-1-1

Retailers and Salesmen (RS) C1-4 txl.Engineers C4-1-2

Students and Learners(SL) C1-5 Publishers C4-2-1

Scientists and Doctors (SD) C2-1 Designers C4-3-1

Scientists and Doctors (SD) C2-1 Labours C5-1-1

Analysts and Policymakers (AP) C2-2 Architects C5-1-2

90 JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011

After generating the snowflake schema, the schema

visualizer displays the diagrammatical view of the

schema in the form of dimension and fact tables along

with the relationships among the tables. The snowflake

schema view, generated for the census data, is shown in

Figure 6.

In addition to this visualization of the schema in the

schema viewer, an automated database is also created in

MS SQL Server 2000. We named the database as

“snowflake_db”. This database has a single fact table and

three dimension tables. The dimension tables are

normalized to form the snowflake schema. Unlike the star

schema, which was generated for the previous

experiment, the snowflake schema has only one

dimension table that is connected directly to the fact

table. All other dimension tables are connected to form a

normalize relationship.

After the database has been created, a separated OLAP

database is created; this database will be connected to the

database server to get the source data. The source data is

the actual clustered database which is generated

automatically. We named the OLAP database as

“Census_OLAP_DB” and proceeded to the next step of

this experiment which is the cube construction process.

The same cube construction process is followed as done

in the previous experiment using the cube wizard of MS

Analysis Services. The difference in this experiment is

that we selected “no_of_persons” as a measure and

defined the dimensional hierarchy in a different way. For

the snowflake schema, we selected the 3 dimension tables

and created a dimensional hierarchy.

The dimension table named “dim_parent” is connected to

the “dim_child” dimension table which is further

connected with the “dim_child_child” table. This way we

can achieve a single dimensional view of the data in

which we selected the hierarchy of that single dimension

from all these tables that are connected in a normalized

way. Snowflake schema has this advantage of selecting

the levels of a dimension form various dimension tables.

Therefore, customized levels for a single dimension were

selected to build a “dim_cluster” for the cube. The

dimension created for the cube can be visualized at

different levels using the dimension browser window.

After defining the dimensions and measures, the cube

storage type has been set as MOLAP in the same way as

done in the previous experiment using the storage design

wizard. Finally, the cube with all these settings is

processed and constructed. We visualize the cube data

using the front-end prototype which has an embedded

visualization enhancement API that the Dundas OLAP

services. This front-end data visualization prototype

connects with the OLAP server as explained in the

previous experiment and displays the cube data in a rich

graphical manner. In the end, we discuss some of the

results achieved through this experiment. The main

objective of the experiment was to emphasize the targeted

and focused analysis. We explain this using a few

problems.

Figure 6. Automatically generated snowflake schema

JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011 91

• Problem : Find out which individual industrial

group of people that is dominant in the census

dataset, furthermore identify the occupation

group which has the largest value and state the

number of people and the exact occupation title.

To solve this, we can easily identify the cluster which has

the maximum number of people as shown in Figure 7. It

is evident from the above figure that C1 cluster has the

maximum number of records, so the focus for the

analysis is only C1 Cluster. We visually explore C1 and

the results of exploration are given in Figure 8.

Again, it is easily recognized that C1 has a specific sub

cluster called “Teachers and Trainers (TT)” which has the

maximum number of people in it. We further drill down

into the sub-cluster TT to find out the exact occupation

which has the maximum number of people. Figure 9

displays the drill down change after clicking the “+” sign

on the chart.

It is evident from the drill down result that in the cluster

named Teachers and Trainers (TT), Sr.Teachers are the

maximum in number; there are 4112 senior teachers in

the dataset. This was a problem which has been easily

solved with quick visual exploration of data. In order to

retrieve such information without the presence of

integrated OLAM architecture, it would have taken a lot

of time, firstly, to construct a query to retrieve such

result. Secondly, the query has to be processed against the

whole data set which is very huge, so it will increase the

query response time as well.

Figure 7. Chart view showing total record in each cluster

Figure 8. Chart view of sub cluster C1

92 JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011

Figure 9. Drill down operation on sub cluster of C1

Thirdly, such rich visual exploration of data using charts

and graphs is also not possible with simple querying and

retrieval of data.

It can be seen from the results of the above two

experiments that integrated OLAM architecture can

greatly enhance the analytical power of the existing

system. It not only automates the schema generation

process but also integrates the enhanced OLAP

functionalities with data mining techniques. When all

these aspects are combined in an efficient way, the

problem solving ability of the OLAM systems is

increased significantly. In the next section, we present a

comparison of the proposed OLAM architecture with the

previous work done in the past regarding OLAM

technology to highlight the contribution and the benefits

achieved through the experimentation.

B. Comparison and discussion

In this section, we compare the proposed architecture

with a number of other OLAM architectures in order to

highlight the significance of this paper work and to

discuss the achievements of this paper. The proposed

integrated OLAM architecture consists of three major

aspects which are 1) Data mining (hierarchical

clustering), 2) Automatic Schema generation (clustered

data) and 3) Enhanced OLAP (performance improvement

& visualization enhancement). Table 2 depicts the major

contribution in the area of OLAM.

TABLE 2

OLAM ARCHITECUTE COMPARISON TABLE

Goil et al. High performance OLAP and Data

Mining on parallel computers

Asghar et al. Enhancing OLAP functionality

using self-organizing neu ral

networks

Messaoud et al. A new OLAP aggregation based

on AHC technique

J. Han Towards online analytical mining

in large databases

Z. Hua Online Analytical Mining of

Association Rules

Joseph et al. Online Analytical Mining Web

Pages-Tick sequence

Dzeroski et al.

Using Data Mining and OLAP to

Discover Patterns in a Database

of Patients with Y-Chromosome

Deletions

Dehne et al. Coarse grained parallel Online

Analytical Processing for data

mining

PROPOSED WORK Integrated OLAM for enhanced

visualization and target ed analysis

AUTHORS WORK TITLE Data Mining

(Hierarchical

Clustering)

Enhanced

OLAP

(Performance &

Visualization)

Automatic

Schema

Generation

JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011 93

It is evident from the table below that the previous work

was more intended towards typical OLAM as compared

to the proposed work which gives direction towards an

integrated OLAM system.

Therefore, the proposed work is significant as the

integrated OLAM architecture can be easily implemented

and used for targeted and focused analysis. Now, we

discuss the major achievements gained through the

architectural implementation. We start with data mining.

• Data mining (hierarchical clustering): We

used the HCE tool for the hierarchical clustering

of data. The major benefit of using this

clustering tool is that it allows users to perform

different types of hierarchical clustering. It has

almost complete support for different kinds of

hierarchical clustering algorithms. It means that

at the mining layer, user can select the algorithm

of his/her choice among all the available

algorithms. Furthermore, the linkage method and

clustering direction can also be given according

to user preference. The accessibility of these

clustering algorithms and parameters supports

the OLAM concept as it allows selection of

various mining algorithms for hierarchical

clustering.

• Automatic schema generation: In addition to

the above mentioned powerful capability of

selecting the hierarchal clustering algorithm, the

proposed architectural implementation makes it

more constructive and dominant by reading the

clustering results and automatically generating

OLAP schema from it. The automatic schema

builder allows three types of schema generation

which include star, snowflake and galaxy. The

implementation provides direction for actually

creating an automated database which not only

holds the schema structure but also uploads the

data from clustered results into specific

dimension and fact tables. This is a very major

achievement, as it reduces all the time and effort

which is required to do the extract, transform

and load (ETL) process in the data warehouse

environment. It means that once you have

generated a schema of a particular type then you

do not have to worry about the data population

process. The automatically generated database is

always ready to become the data source for the

cube construction phase in the OLAP server.

• Enhanced OLAP: As the proposed work

emphasize the integration of enhanced OLAP

system in an integrated OLAM architectural

implementation. We achieved the OLAP

enhancement in terms of its performance and

visualization. Firstly, we discuss the

performance improvement achieved through the

proposed work. We have used ForestCoverType

dataset in the first experiment. But, to calculate

the cube construction time, we created a manual

star schema using SQL queries from the dataset

without passing it to the hierarchical clustering

tool. This non-clustered data set became the

source for the cube in the OLAP server. We

further constructed the ForestCube have 3

dimension tables and 1 fact table. The

ForestCube process time has been noted and it

was observed that it took 1 second to process

60180 non clustered rows of data to construct

the cube completely. The process time and the

rows calculated are shown in Figure 9.

Figure 10. Cube processing time of the data cube

94 JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011

Time Comparison

200

400

600

800

1000

1200

1 3 5 7 9 1113151719

olume of Data (00)

Processi ng Ti me

Non cluster Cube Processing time in (00) sec

Cluster Process ing time in sec

Figure 11. Cube processing time and data size comparison graph

Similarly, we noted the cube construction time for the

clustered ForestCoverType data, which came as a result,

through hierarchical clustering tool, and found that this

clustered data cube was instantly processed and its

construction time was less than a second. Figure 10

shows the cube construction time comparison of both

clustered and non clustered dataset.

It is evident from the time comparison graph that the cube

processing time can be decreased by using the clustered

data and the user can perform the targeted and fast

multidimensional analysis on the clustered cube. Hence,

it validates the proposed architecture for the performance

improvement of OLAP data cube in terms of this reduced

computational time. The second aspect of the enhanced

OLAP is the visualization capability; Figure 11 shows the

outputs of the ForestCube in various visual chart formats.

With the development of the front-end analysis prototype

and embedding the OLAP data visualization controls of

Dundas Software, we provided the OLAP user with a rich

user interface to perform targeted and interactive visual

exploration of the data. User can view the same data in a

number of charts and graphs just by selecting the chart

type from the toolbar present in the developed prototype.

It is clear from the above figure that the visualization of

cube data has been enhanced to facilitate the analyst to

visually explore the hierarchical data and to perform the

basic OLAP operation efficiently and quickly.

In this section, we presented the experiments done on two

real life data sets. Furthermore, we discussed the

experimental results and validated the proposed

architecture by evaluating the results. Finally, we

addressed the aims and objectives which are achieved

through the proposed integrated OLAM architecture

implementation. Experimental results prove that the

proposed work in this paper intensify the focused and

targeted analysis of data and supports the overall

framework of OLAM in a very significant and efficient

way.

Figure 12. Preview of various chart types in the developed prototype

JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011 95

VII. CONCLUSION AND FUTURE WORK

The research conducted in this paper was motivated by

two observations. At one side, OLAM (combination of

OLAP and Data mining), is gaining momentum as an

imminent technology for decision support. On the other

hand, integration of enhanced OLAP systems with any

data mining technique is missing. This gap of integration

limits the analysis capabilities of the standard OLAM

architecture. The proposed work presents an attempt to

reduce the gap of integration by proposing an integrated

OLAM architecture. In addition to reduce the gap, the

proposed architecture also allows automation in the

process of OLAP schema generation. The purpose of this

addition is to intensify the proposed architecture to deal

with the new challenges imposed by emerging decision

support applications.

The results presented in this paper improve the existing

OLAM architectures by overcoming the existing

limitations and introducing the automatic schema

generation layer in it. Due to the limited timeframe of our

work we deliberately excluded some properties of data

mining techniques as they would have expanded the

scope of the resulting architecture. Our work can be

continued in research areas, such as integration of OLAP

and data mining, automatic schema generation, enhanced

OLAM systems or by intensifying the areas central to this

paper.

Firstly, we have only considered a particular data mining

technique which is hierarchical clustering. Future work

can be the use of other mining techniques such as

association rule mining, classification, predication and

others with enhanced OLAP systems to benefit the

existing OLAM architecture. Secondly, the automatic

schema generation layer in our architectures is only

useful for the hierarchal clustering results which have a

parent-child relationship to generate schemas of different

types. This area can also be improved for the different

types of mined results as different data mining techniques

produce different mining results. We are exploring other

data mining techniques that can also be integrated with

the enhanced OLAP system and also the mined results

can be used to automatically build schema of different

types. Another direction for future work is that we intend

to enhance visualization capabilities of the hierarchical

clustering results so that the clusters should be named

properly for identification purpose.

REFERENCES

[1]. S. Chaudhuri and U. Dayal, An overview of data

warehousing and OLAP technology, ACM SIGMOD

Record, Vol. 26 (1997), pp. 65-74.

[2]. A. Cuzzocrea, D. Sacca and P. Serafino, A hierarchy

driven compression technique for advanced OLAP

visualization of multidimensional data cubes, in Proc. of

8th Int’l Conf. on Data Warehousing and Knowledge

Discovery (DaWak), (Springer Verlag 2006), pp. 106-119

[3]. S. Maniatis, P. Vassiliadis, S. Skiadopoulos and Y.

Vassiliou, Advanced visualization for OLAP, in Proc. of

6th ACM Int’l Workshop on Data Warehousing and OLAP

(DOLAP), (ACM New York, 2003), pp. 9-16.

[4]. S. Mansmann and M. Scholl, Exploring OLAP aggregates

with hierarchical visualization techniques, in Proc. of

ACM Symposium on Applied Computing (2007), pp.

1067-1073.

[5]. S. Goil and A. Choudhary, “High performance OLAP and

data mining on parallel computers,” Data Mining and

Knowledge Discovery, vol. 1, no. 4, pp. 391-417, Dec.

1997.

[6]. D. Papadias, P. Kalnis, J. Zhang and Y. Tao, “Efficient

OLAP operations in spatial data warehouse,” in Proc. of

the 7th Int’l Symposium on Advances in Spatial and

Temporal Databases (SSTD), Springer-Verlag , 2001, pp.

443-459

[7]. S. Asghar, D. Alahakoon and A. Hsu, “Enhancing OLAP

functionality using self-organizing neural networks,”

Neural, Parallel and Scientific Computations, vol. 12, no.

1, pp. 1-20, March 2004.

[8]. R. B. Messaoud, O. Boussaid and S. Rabaseda, “A new

OLAP aggregation based on the AHC technique,” in Proc.

of the 7th ACM Int’l Workshop on Data Warehousing and

OLAP (DOLAP), ACM New York, 2004, pp. 65-72.

[9]. S. Maniatis, P. Vassiliadis, S. Skiadopoulos and Y.

Vassiliou, “Advanced visualization for OLAP,” in Proc. of

the 6th ACM Int’l Workshop on Data Warehousing and

OLAP (DOLAP), ACM New York, 2003, pp. 9-16.

[10]. M. Sifer, “A visual interface technique for exploring

OLAP data with coordinated dimension hierarchiches,” in

Proc. of the 12th ACM Int’l Conf. on Information and

Knowledge Management (CIKM), ACM New York, 2003,

pp. 532-535.

[11]. A. Voss, V. Hernandez, H. Voss and S. Scheider,

“Interactive visual exploration of multidimensional data:

requirements for common GIS with OLAP,” in Proc. of the

15th Int’l Workshop on Database and Expert Systems

Applications (DEXA), 2004, pp. 883-887.

[12]. M. Scotch and B. Paramanto, “SOVAT: Spatial OLAP

visualization and analysis tool,” in Proc. of the 38th

Annual Hawaii Int’l Conf. on Systems Sciences (HICSS),

2005, p. 165.

[13]. A. Cuzzocrea, D. Sacca and P. Serafino, “A hierarchy

driven compression technique for advanced OLAP

visualization of multidimensional data cubes,” in Proc. of

the 8th Int’l Conf. on Data Warehousing and Knowledge

Discovery (DaWak), Springer Verlag, 2006, pp. 106-119.

[14]. K. Techapichetvanich and A. Datta, “Interactive

visualization for OLAP,” in Int’l Conf. on

Computational Science and its Applications (ICCSA),

2005, pp. 206-214.

[15]. S. Mansmann and M.Scholl, “Extending visual OLAP for

handling irregular dimensional hierarchies,” in Proc.

of 8th Int’l Conf. on Data Warehousing and

Knowledge Discovery (DaWaK'06), 2006, pp. 95-105.

[16]. S. Mansmann and M. Scholl, “Visual OLAP: A new

paradigm for exploring multidimensional aggregates,”

in Proc. of IADIS Int’l Conf. on Computer Graphics

and Visualization (CGV), 2008, pp. 59-66.

[17]. S. Asghar and M. Usman, “Enhancing OLAP performance

and visualization,” in Proc. of 6th Int’l Conf. on

96 JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011

Information Science Technology and Management

(CISTM), 2008, pp. 59-70.

[18]. Fayyad U.M, Piatesky-Shapiro G., Smyth p. and

Uthurusamy R., “From data mining to knowledge

discovery: An overview,”in Proc. of Advances in data

mining and knowledge discovery, MIT Press, 1996,

pp. 1-34.

[19]. J. Hipp, U. Guentzer and G. Nakhaeizadeh,

“Algorithms for association ruling mining – a general

survey and comparison,” ACM SIGKDD Explorations

Newsletter, vol. 2, no. 1, pp. 58 – 64, June 2000.

[20]. R. C. Holte, “Very simple classification rules perform

well on most commonly used datasets,” Machine

Learning, vol. 11, no. 1, pp. 63-90, April 1993.

[21]. A. K. Jain, M. N. Murty and P. J. Flynn, “Data

clustering: A review,” ACM Computing surveys

(CSUR), vol. 31, no. 3, pp. 264-323, Sep. 1999.

[22]. M. J. Shaw, C. Subramaniam, G. W. Tan and M. E.

Welge, “Knowledge management and data mining for

marketing,” Decision Support Systems, vol. 31, no. 1,

pp. 127-137, May 2001.

[23]. J. Han, “Towards online analytical mining in large

databases,” ACM SIGMOD Record, vol. 27, no. 1, pp.

97-107, March 1998.

[24]. J. Han, S. H. S. Chee and J. Y. Chiang, “Issues for on-

line analytical mining of data warehouses,” in Proc. of

the SIGMOND Workshop on Research Issues on Data

Mining and Knowledge Discovery (DMKD), Seattle,

1998, pp. 2:1-2:5.

[25]. H. Zhu, “Online analytical mining of association

rules,” Master Thesis, Simon Fraser University, 1998,

pp. 1-117.

[26]. J. Fong, H. K. Wong and A. Fong, “Online analytical

mining Web-pages tick sequences,” J. of Data

Warehousing, vol. 5, no. 4, pp. 59-67, 2000.

[27]. S. Dzeroski, D. Hristovski and B. Peterlin, “Using

data mining and OLAP to discover patterns in a

database of patients with Y chromosome deletions,”

in Proc. AMIA Symp., 2000, pp. 215–219.

[28]. F. Dehne, T. Eavis and A. Rau-Chaplin, “Coarse

grained parallel on-line analytical processin (OLAP)

for data mining, in Proc. of the Int’l Conf. on

Computational Science (ICCS), 2001, 589-598.

[29]. V. Markl, F. Ramasak and R. Bayer, “Improving

OLAP performance by multi-dimensional hierarchical

clustering,” in Proc. of the 1999 Int’l Symposium on

Database Engineering and Applications (IDEAS ),

1999, p. 165.

[30]. V. Markl and R. Bayer, “Processing relational OLAP

queries with UB-trees and multidimensional

hierarchical clustering,” in Proc. of the Int’l.

Workshop on Design and Management of Data

Warehouses (DMDW), 2000, pp. 1-10.

[31]. N. Karayannidis, T. Sellis and Y. Kouvaras, “CUBE

file: A file structure for hierarchically clustered OLAP

cubes,” Advances in Database Technology, LNCS,

Springer Verlag, Berlin-Heidelberg, pp. 621-638,

2004.

[32]. D. Theodoratos and A. Tsois, “Heuristic optimization

of OLAP queries in multidimensionally hierarchically

clustered databases,” in Proc. of the 4th ACM Int’l

Workshop on Data Warehousing and OLAP

(DOLAP), ACM New York, 2001, pp. 48-55.

[33]. K. Hann, C. Sapia and M. Balaschka, “Automatically

generating OLAP schemata from conceptual graphical

models,” in Proc. of the 3rd ACM Int’l. Workshop on

Data Warehousing and OLAP (DOLAP), ACM New

York, 2000, pp. 9-16.

[34]. V. Peralta, A. Marotta and R. Ruggia, “Towards the

automation of data warehouse design,” Technical

Report TR-03-09, InCo, Universidad de la República,

Montevideo, Uruguay, June 2003.

[35]. N. Tryfona, F. Busborg and J. G. B. Christiansen

“StarER: A conceptual model for data warehouse

design,” Proc. of the 2nd ACM Int’l. Workshop on

Data Warehousing and OLAP (DOLAP), ACM New

York, 1999, pp. 3-8.

[36]. Y. Song, et. al., “SAMSTAR: An automatic tool for

generating star schemas from an entity-relationship

diagram,” in Proc. of the 27th Int’l. Conf. on

Conceptual Modeling, LNCS 2008,vol. 5231 pp. 522-

523.

[37]. S. R. Maddi and V. Khan, “Comparative analysis of

on-line analytical processing tools,” University essay

from IT-universitetet I Göteborg, Sweden 2007.

[38]. M. Reed, “A definition of data warehousing,”

[Onlline]. Available:

http://www.intranetjournal.com/features/datawarehou

sing.html. [Accessed: April 28, 2009].

[39]. R. Kimball and M. Ross, The Data Warehouse

Toolkit: The Complete Guide to Dimensional

Modeling, 2nd ed., New York: John Wiley, 1996.

[40]. J. Widom, “Research problems in data warehousing,”

in Proc. of the 4th Intl. Conf. on Information and

Knowledge Management (CIKM), 1995, pp. 25-30.

[41]. S. Chaudhuri and U. Dayal, “Data warehousing and

OLAP for decision support,” in Proc. of the ACM Intl.

Conf. on Management of Data, ACM New York,

1997, pp. 507-508.

[42]. G. Colliat, “OLAP, relational, and multidimensional

database systems,” ACM SIGMOND Record, vol. 25,

no. 3, pp. 64-69, 1996.

[43]. H. Gupta, V. Harinarayan, A. Rajaraman and J. D.

Ullman, “Index selection for OLAP,” in Proc. of the

13th Intl. Conf. on Data Engineering (ICDE), 1997,

pp. 208-219.

[44]. W. H. Inmon, “The data warehouse and data mining,”

Communications of the ACM, vol. 39, no. 11, pp. 49-

50, 1996.

[45]. ADSS OLAP Help. [Online]. Available:

http://www.agroict-olap.org/OLAP_Help.aspx.

[Accessed: April 28, 2009].

[46]. T. J. Teorey, „Data warehousing, OLAP, and data

mining, in Database Modeling and Design, 3rd ed. ,

Morgan Kaufmann, 1999, pp. 211-250.

[47]. IBM Journal of Research and Development. [Online].

Available:

https://www.research.ibm.com/journal/sj/414/colossi.

html. [Accessed: April 28, 2009].

[48]. E. Thomsen, “OLAP Solutions: Building

Multidimensional Information Systems,” 2

nd . New

York: John Wiley, 2002.

[49]. B. Dinter and C. Sapia et. al., "OLAP marketing and

research: Initiating the cooperation," J. of Computer

Science and Information Management, vol. 2, no. 3,

1999.

[50]. M S. Chen, J. Han and P. S. Yu, “Data mining: An

overview from a database perspective,” IEEE

Transactions on Knowledge and Data Engineering,

vol. 8, no. 6, pp. 866-883, Dec. 1996.

[51]. U. Fayyad, G. Piatetsky-Shapiro and P. Smyth,

“Knowledge discovery and data mining: Towards a

unifying framework,” in Proc. of the 2nd Intl. Conf. on

JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011 97

Knowledge Discovery and Data Mining, 1996, pp. 82-

88.

[52]. D. Hand, H. Mannila and P. Smyth, “Principles of

Data Mining,” [Online]. Available:

http://cat.inist.fr/?aModele=afficheN&cpsidt=189586

26. [Accessed: April 28, 2009].

[53]. U. Fayyad, G. Piatetsky-Shapiro and P. Smyth, “From

data mining to knowledge discovery in databases,” AI

Magazine, 1996.

[54]. P. Berkhin, “A survey of clustering data mining

techniques,” in Grouping Multidimensional Data,

Springer Berlin-Heidelberg, 2006, pp. 25-71.

[55]. A. Berson, S. Smith and K. Thearling, “An Overview

of Data Mining Techniques,” [Online]. Available:

http://www.thearling.com/text/dmtechniques/dmtechn

iques.htm. [Accessed: April 28, 2009].

[56]. D. Beeferman and A. Berger, “Agglomerative

clustering of a search engine query log,” in Proc. of

the 6th Intl. Conf. on Knowledge Discovery and Data

Mining ACM SIGKDD, 2000, pp. 407-416.

[57]. S. O. Deininger, et. al., “MALDI imaging combined

with hierarchical clustering as a new tool for the

interpretation of complex human cancers,” J.

Proteome Res., vol. 7, no. 12, pp. 5230-5236, 2008.

[58]. C. Li and W. H. Wong, “DNA-Chip Analyzer

(dChip),” [Online]. Available:

http://biosun1.harvard.edu/complab/dchip/Li%2003.p

df. [Accessed: April 28, 2009].

[59]. V. Hatzivassiloglou, et. al., “SIMFINDER: A flexible

clustering tool for summarization,” in NAACL

Workshop on Automatic Summarization, 2001, pp. 41-

49.

[60]. J. Seo, et. al., “Interactive color mosaic and

dendrogram displays for signal/noise optimization in

microarray data analysis,” in Proc. of the Intl. Conf.

on Multimedia and Expo-Volume 3, 2003, pp. 461-

464.

[61]. R. Rak and W. Stach, “Evaluation of XLMiner demo

version 2,” report, 2004. [Online]. Available:

http://www.cs.ualberta.ca/~zaiane/courses/cmput695-

04/work/A2-reports/xlminer.pdf. [Accessed: April 28,

2009].

[62]. [Online]. Available: http://www.cs.umd.edu/hcil/hce/.

[Accessed: April 28, 2009].

[63]. Netz, S. Chaudhuri, J. Bernhardt and U. Fayyad,

“Integration of data mining and relational databases,”

in Proc. of the 26th Intl. Conf. on Very Large

Database, 2000, pp. 719-722.

[64]. [Online].

Available:http://www.ibm.com/developerworks/ration

al/library/07/0626_johnson/index.html. [Accessed:

April 28, 2009].

[65]. D. L. K. Johnson, “Reporting with rational portfolio

manager version 7.1,” June 26, 2007. [Online].

Available:

http://www.ibm.com/developerworks/rational/library/

07/0626_johnson/index.html [Accessed: April 28,

2009].

[66]. Stanford Technology Group, “Designing the Data

Warehouse on Relational Databases,” 1996. [Online].

Available:

http://www.informix.com.ua/articles/rolap/rolap.htm.

[Accessed: April 28, 2009].

[67]. D. L. Moody and M. A.R. Kortink, “From enterprise

models to dimensional models: A methodology for

data warehouse and data mart design,” in Proc. of the

Intl. Workshop on Design and Management of Data

Warehouses (DMDW), 2000, pp. 5:1-5:12.

[68]. “Data Warehouse Design,“ [Online]. Available:

http://www.pafis.shh.fi/%7Eargarg03/SFIS/WS1/data

w_design.htm. [Accessed: April 28, 2009].

[69]. S. Asghar, D. Alahakoon and D. Taniar, “Enhancing

OLAP functionality using association rules,” in Proc.

of 2nd Int’l Conf. on Computational Intelligence,

Robotics and Autonomous Systems (CIRAS ), 2003,

pp. 1-6.

[70]. A. Rauber and P. Tomsich, “An architecture for

modular OLAP Systems: Supporting distributed and

parallel query processing using cooperating CORBA

objects,” in 10th Int’l Workshop on Database and

Expert Systems Applications (DEXA), 1999, p. 45.

[71]. S. Asghar and M. Usman, “Integrated enhancement of

OLAP performance and visualization using self

growing organizing neural networks,” in Int’l J. of

Information Tech. and Decision Making, 2009,

submitted.

[72]. Available:http://support.dundas.com/OnlineDocument

ation/WebOLAP/Overview%20of%20Dundas%20OL

AP%20Architecture.html. [Online]. [Accessed: Apr.

2009].

[73]. J. A. Blackard, D. J. Dean and C. W. Anderson,

Forest cover type, The UCI KDD Archive

[http://kdd.ics.uci.edu]. Irvine, CA: University of

California, Department of Information and Computer

Science (1998).

[74]. C. Meek, B. Thiesson, D.Heckerman,, US Census

Data (1990), The UCI KDD Archive

[http://archive.ics.uci.edu/ml/datasets/US+Census+Da

ta 281990 29], US Department of Commerce.

[75]. Available.

http://archive.ics.uci.edu/ml/databases/census1990/US

Census1990raw.html. Online [Accessed: Apr. 2009].

[76]. Available.

http://archive.ics.uci.edu/ml/databases/census1990/US

Census1990raw.coding.htm Online [Accessed: Apr.

2009].

[77]. Available.

http://archive.ics.uci.edu/ml/databases/census1990/US

Census1990.mapping.sql Online [Accessed: Apr.

2009]

Muhammad Usman received BS

computer science degree from Comsats

institute of information technology,

Pakistan in 2006 and MS software

engineering degree from Shaheed

Zulfikar Ali Bhutto institute of science

and technology, Pakistan in 2009.

He has worked as a Data warehouse

developer in the center of Agro-

informatics research (CAIR), Pakistan. He has published 5

conference paper and 1 journal article. He is currently

researching in data mining, online analytical processing,

decision support systems, knowledge discovery, neural

networking, pattern recognition and soft computing.

Mr. Usman is a recipient of National Agriculture and IT

advancement award in recognition of his work in Agriculture

decision support system from FAST national university,

Islamabad, Pakistan. Now he is studying a PhD program at

Auckland University of Technology, New Zealand.

98 JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011

Sohail Asghar is an Associate Professor

of Computing, Faculty of Computer

Science, Mohammad Ali Jinnah

University, Islamabad, Pakistan. In 1994,

he graduated with honors in Computer

Science from the University of Wales,

United Kingdom. From 1994 to 2002, he

worked as a Senior Software Engineer in

a software company in Islamabad. He

then obtained his PhD in Information Technology at Monash

University, Melbourne Australia in 2006.

He has taught and researched in the Data Mining and Decision

Support Systems areas, and he has published extensively in both

Australian and international journals as well as conferences

proceedings. He has also consulted widely on information

Technology matters, especially within the framework of disaster

management systems. In 2004 he won the Australian Post

Graduate Award for Industry.

Dr. Asghar is a member of, the Australian Computer Society

(ACS), and IEEE.

JOURNAL OF EMERGING TECHNOLOGIES IN WEB INTELLIGENCE, VOL. 3, NO. 2, MAY 2011 99

Data Mining Techniques in Database Systems

Article

Full-text available

Feb 2017

Ledion Lico

At the current stage the technologies for generating and collecting data have been advancing rapidly. The main problem is the extraction of valuable and accurate information from large data sets. One of the main techniques for solving this problem is Data Mining. Data mining (DM) is the process of identification and extraction of useful information in typically large databases. DM aims to automatically discover the knowledge that is not easily perceivable. It uses statistical analysis and artificial intelligence (AI) techniques together to address the issues. There are different types of tasks associated to data mining process. Each task can be thought of as a particular kind of problem to be solved by a data mining algorithm. The main types of tasks performed by DM algorithms are: Classification, Association, Clustering, Regression, Anomaly Detection, Feature Extraction, Time Series Analyses.

Production Planning and Scheduling Using Machine Learning and Data Science Processes

Chapter

Full-text available

Dec 2020

Increasing manufacturing efficiency has been a constant challenge since the First Industrial Revolution. What started as mechanization and turned into electricity-driven operations has experienced the power of digitalization. Currently, the manufacturing industry is experiencing an exponential increase in data availability, but it is essential to deal with the complexity and dynamics involved to improve manufacturing indicators. The aim of this study is to identify and allow an understanding of the unfilled gaps and the opportunities regarding production scheduling using machine learning and data science processes. In order to accomplish these goals, the current study was based on the Knowledge Development Process – Constructivist (ProKnow-C) methodology. Firstly, selecting 30 articles from 3608 published articles across five databases between 2015 and 2019 created a bibliographic portfolio. Secondly, a bibliometric analysis, which generated comparative charts of the journals’ relevance regarding its impact factor, scientific recognition of the articles, publishing year, highlighted authors and keywords was carried out. Thirdly, the selected articles were read thoroughly through a systemic analysis in order to identify research problems, proposed solutions, and unfilled gaps. Then, research opportunities identified were: (i) Big data and associated analytics; (ii) Collaboration between different disciplines; (iii) Solution Customization; and (iv) Digital twin development.

Modified Approach for Classifying Multi- Dimensional Data-Cube Through Association Rule Mining for Granting Loans in Bank

Article

Mar 2016
IJETTCS

Dr. K Kavitha

Enterprise Architecture Model that Enables to Search for Patterns of Statistical Information

Article

Full-text available

Jun 2013
IJARAI

Oleg Chertov

Enterprise architecture is the stem from which developing of any departmental information system should grow and around which it should revolve. In the paper, a fragment of an enterprise architecture model is built using ArchiMate language. This fragment enables to search for information in enterprises which do not work in productive industry. Such enterprises include official statistics. The proposed model embraces all three architectural levels of corresponding information systems, namely, OLTP, OLAP, and Data Mining. Particularly, the latter level enables to search for patterns of statistical information.

Analyzing Customer Behavior Using Online Analytical Mining (OLAM)

Chapter

Jan 2015

Thanachart Ritbumroong

Online Analytical Mining (OLAM) is an architecture integrating data mining into OLAP. With this integration, data mining algorithms can be performed with OLAP abilities. OLAM enables users to choose a particular portion of data and analyze them with data mining models. Previous studies have provided examples of OLAM applications with the motivation to improve technical performance. This chapter reviews the capabilities of OLAM and discusses the well-known concept encompassing the analysis of customer behavior. The underlying motivation of this chapter is to present the opportunities for the development of OLAM to support the customer behavior analysis. Three main directions of the advancement in OLAM are proposed for future research.

OLAM

Chapter

Jan 2016

Matteo Golfarelli

Analyzing Customer Behavior Using Online Analytical Mining (OLAM)

Article

Jan 2014

Thanachart Ritbumroong

A Reactive Scheduling Strategy Applied On MapReduce OLAM Operators System

Article

Nov 2012

The combination of Data warehousing and data analysis techniques such as OLAP (Online Analytic Processing) and data mining through the Hadoop framework is an innovative way to treat large volumes of data. However, this way poses serious scheduling and combining tasks issues that bring more challenges. In this paper, we propose strategies to answer these questions, namely parallel OLAM (Online Analytic Mining) MapReduce Operators and a Reactive Scheduling Policy. OLAM MapReduce Operators divide jobs into two parts, the first includes all the operators that are used to create an OLAM CUBE and the second includes those who exploit the cube by data mining algorithms. The proposed policy coordinates the workflow generated by these operators, relying on model-based events. Our simulation experience shows that our strategy has a cumulative force that it reduces the execution time of the entire cluster at each request.

Discovering diverse association rules from multidimensional schema

Article

Jun 2013
EXPERT SYST APPL

The integration of data mining techniques with data warehousing is gaining popularity due to the fact that both disciplines complement each other in extracting knowledge from large datasets. However, the majority of approaches focus on applying data mining as a front end technology to mine data warehouses. Surprisingly, little progress has been made in incorporating mining techniques in the design of data warehouses. While methods such as data clustering applied on multidimensional data have been shown to enhance the knowledge discovery process, a number of fundamental issues remain unresolved with respect to the design of multidimensional schema. These relate to automated support for the selection of informative dimension and fact variables in high dimensional and data intensive environments, an activity which may challenge the capabilities of human designers on account of the sheer scale of data volume and variables involved. In this research, we propose a methodology that selects a subset of informative dimension and fact variables from an initial set of candidates. Our experimental results conducted on three real world datasets taken from the UCI machine learning repository show that the knowledge discovered from the schema that we generated was more diverse and informative than the standard approach of mining the original data without the use of our multidimensional structure imposed on it.

Data Guided Approach to Generate Multi-dimensional Schema for Targeted Knowledge Discovery

Conference Paper

Dec 2012

Data mining and data warehousing are two key technologies which have made significant contributions to the field of knowledge discovery in a variety of domains. More recently, the integrated use of traditional data mining techniques such as clustering and pattern recognition with data warehousing technique of Online Analytical Processing (OLAP) have motivated diverse research areas for leveraging knowledge discovery from complex real-world datasets. Recently, a number of such integrated methodologies have been proposed to extract knowledge from datasets but most of these methodologies lack automated and generic methods for schema generation and knowledge extraction. Mostly data analysts need to rely on domain specific knowledge and have to cope with technological constraints in order to discover knowledge from high dimensional datasets. In this paper we present a generic methodology which incorporates semi-automated knowledge extraction methods to provide data-driven assistance towards knowledge discovery. In particular, we provide a method for constructing a binary tree of hierarchical clusters and annotate each node in the tree with significant numeric variables. Additionally, we propose automated methods to rank nominal variables and to generate candidate multidimensional schema with highly significant dimensions. We have performed three case studies on three real-world datasets taken from the UCI machine learning repository in order to validate the generality and applicability of our proposed methodology.

DATA MINING & DATA WAREHOUSING

Book

Full-text available

Jul 2022

Thamizharasi Ramalingam

This book describes the basic concepts of Data mining and Data warehousing concepts

Enhancing OLAP Performance and Visualization

Conference Paper

Full-text available

Aug 2008

SAMSTARplus: An Automatic Tool for Generating Multi-Dimensional Schemas from an Entity-Relationship Diagram

Article

Full-text available

Jan 2009

This paper presents a tool that automatically generates multidimensional schemas for data warehouses from OLTP entity-relationship diagrams (ERDs). Based on user's input parameters, it generates star schemas, snowflake schemas, or a fact constellation schema by taking advantage of only structural information of input ERDs. Hence, SAMSTARplus can help users reduce efforts for designing data warehouses and aids decision making.

From data mining to knowledge discovery: an overview

Conference Paper

Feb 1996

The Data Warehouse Toolkit: The Complete Guide to Dimensional Modeling

Article

Jan 2011
SIGMOD REC

Issues for On-Line Analytical Mining of Data Warehouses

Article

Jan 1998

Knowledge Discovery and Data Mining: Towards a Unifying Framework

Conference Paper

Jan 1996

This paper presents a first step towards a unifying framework for Knowledge Discovery in Databases. We describe finks between data milfing, knowledge dis- covery, and other related fields. We then define the KDD process and basic data mining algorithms, dis- cuss application issues and conclude with an analysis of challenges facing practitioners in the field.

Enhancing OLAP functionality using self-organizing neural networks

Article

Mar 2004

On-Line Analytical Processing (OLAP) has become a popular management decision-making tool due to its user-friendly visualization abilities. With this popularity, user's expectation and demands from existing OLAP systems have increased. The described work extends the capabilities of OLAP with additional functionality by using neural network technology. In addition to the usual visualization capabilities, this new technique provides the user with the opportunity to analyse clusters in the data at different levels of abstraction. The technique used for enhancing OLAP functionality is a model called the Growing Self-Organizing Map (GSOM). The GSOM has been developed as a more flexible data mining friendly feature mapping method over the traditional Self-Organizing Map (SOM). One of the major innovations with the GSOM is the possibility of generating feature maps of different levels of data abstraction using a parameter called the spread factor. This spread factor has been used to develop a hierarchical cluster generation and analysis technique called the Dynamic SOM Tree. These hierarchical clusters facilitate the OLAP user to gain insight and obtain prior knowledge of the data set before performing OLAP operations. In addition, the hierarchical clusters from a Dynamic SOM Tree are used to provide the OLAP user with the ability to visualize and select data clusters at different levels of abstraction for further detailed analysis.

Knowledge management and data mining for marketing

Article

May 2001
DECIS SUPPORT SYST

Due to the proliferation of information systems and technology, businesses increasingly have the capability to accumulate huge amounts of customer data in large databases. However, much of the useful marketing insights into customer characteristics and their purchase patterns are largely hidden and untapped. Current emphasis on customer relationship management makes the marketing function an ideal application area to greatly benefit from the use of data mining tools for decision support. A systematic methodology that uses data mining and knowledge management techniques is proposed to manage the marketing knowledge and support marketing decisions. This methodology can be the basis for enhancing customer relationship management. q 2001 Elsevier Science B.V. All rights reserved.

Principles of Data Mining

Book

Jan 2001

The growing interest in data mining is motivated by a common problem across disciplines: how does one store, access, model, and ultimately describe and understand very large data sets? Historically, different aspects of data mining have been addressed independently by different disciplines. This is the first truly interdisciplinary text on data mining, blending the contributions of information science, computer science, and statistics. The book consists of three sections. The first, foundations, provides a tutorial overview of the principles underlying data mining algorithms and their application. The presentation emphasizes intuition rather than rigor. The second section, data mining algorithms, shows how algorithms are constructed to solve specific problems in a principled manner. The algorithms covered include trees and rules for classification and regression, association rules, belief networks, classical statistical models, nonlinear models such as neural networks, and local "memory-based" models. The third section shows how all of the preceding analysis fits together when applied to real-world data mining problems. Topics include the role of metadata, how to handle missing data, and data preprocessing.

An Architecture for Integrated Online Analytical Mining

Abstract and Figures

Recommended publications

A Seamless Integration of Association Rule Mining with Database Systems

YOUR MARK IS MY DIRT: IMPACT OF EMAIL SIGNATURES ON DECISION MAKING

Deriving Concise Association Rules Efficiently while Extracting Closed Itemsets in Depth-first Manne...

Nachrichtenklassifikation als Komponente in WEBIS