Review on Big Data & Analytics – Concepts, Philosophy, Process and Applications

Abstract

Big Data analytics has been the main focus in all the industries today. It is not overstating that if an enterprise is not using Big Data analytics, it will be a stray and incompetent in their businesses against their Big Data enabled competitors. Big Data analytics enables business to take proactive measure and create a competitive edge in their industry by highlighting the business insights from the past data and trends. The main aim of this review article is to quickly view the cutting-edge and state of art work being done in Big Data analytics area by different industries. Since there is an overwhelming interest from many of the academicians, researchers and practitioners, this review would quickly refresh and emphasize on how Big Data analytics can be adopted with available technologies, frameworks, methods and models to exploit the value of Big Data analytics.

Full text

3
BULGARIAN ACADEMY OF SCIENCES
CYBERNETICS AND INFORMATION TECHNOLOGIES  Volume 17, No 2
Sofia  2017 Print ISSN: 1311-9702; Online ISSN: 1314-4081
DOI: 10.1515/cait-2017-0013
Review on Big Data & Analytics – Concepts, Philosophy, Process
and Applications
Kari Venkatram, Geetha Mary A.
School of Computing Science and Engineering, VIT University, Vellore 632 014, Tamil Nadu, India
E-mails: venkat_kari@yahoo.com geethamary@vit.ac.in
Abstract: Big Data analytics has been the main focus in all the industries today. It is
not overstating that if an enterprise is not using Big Data analytics, it will be a stray
and incompetent in their businesses against their Big Data enabled competitors. Big
Data analytics enables business to take proactive measure and create a competitive
edge in their industry by highlighting the business insights from the past data and
trends. The main aim of this review article is to quickly view the cutting-edge and
state of art work being done in Big Data analytics area by different industries. Since
there is an overwhelming interest from many of the academicians, researchers and
practitioners, this review would quickly refresh and emphasize on how Big Data
analytics can be adopted with available technologies, frameworks, methods and
models to exploit the value of Big Data analytics.
Keywords: Data, Big Data, Hadoop, no sql, analytics.
1. Introduction
Data manifestation has been changed radically over last few years. Today’s data is
much different from the past. Data is becoming amorphous, which means many forms
of data without any proper shape or form. Big Data solutions came forward to handle
such types of data. Big Data is referred as Data Intensive Technology (DIT)[1]. There
are significant learnings from the continuous data generation from both science &
technology as well as quality of the business and lives [2]. Big Data is becoming part
of every aspect of our lives. Technologies such as hadoop, cloud computing and
visualization enable data collection, process & store and visualize data insights [1].
Google search was built based on Big Data considerations and google analytics
helped in arresting the virus spread H1N1 in the USA during virus spread. Big Data
and its analytics will be every part of our life in future, right from the schooling, i.e.,
identifying the right school for your kids, helping human to monitor their health
suggesting the remedies for wellbeing and it may recommend you for right partner
based on your interest and many more.
Unauthenticated
Download Date | 7/5/17 9:02 AM

4
2. Evolution of Big Data technology
From the ancient times human knows how to store and process data, the initial storage
being on stones of caves and tamarapatra (leaves), etc., which was referenced by next
generations to interpret the previous generation’s wisdom and heritage. After
invention of Paper around 220AD the data was stored in bunch of papers, which was
nothing but a book. However, the ability to analyse the data is limited to one’s own
brain and knowledge inferred by an individual, which was stepping stone for
inventing an alternate for human brain that is computer evolution. This era began
with punch cards and OMR for data storage.
Since the computer became integral part of every business there was a need to
communicate among computers in many geographical locations, which led to
invention of Internet, which triggered massive improvement in data storage,
processing and analysing techniques that started digital electronics era. In order to
manage and process the data, files systems and database management systems have
been developed. File systems enable to store and access hard disk and DBMS enables
to store and access data. Data generation rate has been tremendously increased from
last two decades and it is the origin for new horizons in data storing and processing
requirements for wide variety, huge volumes of data from kilobytes to zeta bytes.
2.1. Data transformation to wisdom
Information, knowledge and wisdom extracted from the data. Data [3, 4] is nothing
but gathering some of the facts, specifications, objectives, particulars, some details,
statistics, figures, images, audio and video, etc. Information [3, 4] is something
derived after analysing particular data that is being used for taking some of the
business decisions. It is derived by synthesizing and contextualizing the data to
provide value. Knowledge [3, 4] is something we get more understanding and
expertize from the information on some subject. Wisdom [4] is a technique to utilize
particular knowledge at some situation. This is also termed as solution pattern for a
particular problem.
2.2. Data storage evolution
Data storage and retrieval is being controlled by the operating system using file
operations in the underlying file system. Different operating systems follow different
approaches to manage their files in the storage devices. Later relational data base
management systems evolved to address the redundancy issues in the traditional
database management systems by means of table. RDBMS has been widely used in
many of the business applications for data storage and processing. It has been a
widespread across the enterprises. However, it has a limitation to work with
heterogeneous systems as it can support only limited data types and hence there is a
need for custom data types. To overcome this issue a new data base management
systems called object relational database management system has been invented. It
enables to define custom / own data types and methods. ORDBMS support data types
such audios, videos, images and any custom data structures.
Unauthenticated
Download Date | 7/5/17 9:02 AM

5
2.3. Attributes of data in five C’s [3]
As described in the Fig. 1, data has five important attributes. We call them simply 5
Cs such as Clean, Consistent, Confirmed, Current and Comprehensive. All the
attributes are briefly defined in the following paragraph.
Clean: Data in raw form requires cleansing to get into a desired format to gain
some information from it. Consistent: Data to be represented in a way that is
consistent in its representation, so management can take a decision without having
any arguments among the teams. Confirmed: Data to be confirmed to adapted
dimensions across the data and should be consistent to be utilized by any team.
Current: Data should be accurate at that point of time to take some business
decisions. Comprehensive: Data should be inclusive of all the required items from its
granular level from all the sources to take some decision.
Fig. 1. Attributes of data
2.4. Data processing evolution
Storing and processing of large data is a challenge, there are some innovations done
to address these technical and processing challenges. Ex: Grid computing handles
voluminous data, cloud computing deals with high velocity and huge data volumes,
open sourcing technologies for cost reduction and virtualization reduced time to test,
deploy and improves processing speeds, etc. However, there are issues with these
solutions like grids are expensive, clouds are seeming to be slow, open source less
robust, virtualization tends to slowdown the execution process. So, there is a need for
new solution to handle the data challenges.
As depicted in Fig. 2, ERP generates limited data in some megabytes through a
form of purchases and payments details, etc. CRM generates more data in some of
gigabytes and process data about customers, relationships, offers and other details
like segmentation etc. ERP and CRM applications data managed using OLTP. With
the evolution of web technologies lots of data shared among a web servers, hence
data generation increased significantly through web logs. At the same time data
analytics requires historical information to extract data insights. Hence many of
terabytes data were generated. Also, there is a new requirement for storing the
historical data in OLAPs apart from OLTP, which manages transactional data. In the
last few years data has exponentially grown due to IOT (Internet Of Things) [5],
Clean
Consistent
ConfirmedCurrent
Comprehensive
Unauthenticated
Download Date | 7/5/17 9:02 AM

6
mobile apps, sensor data, etc., which requires petabytes of data. Hence there is a need
of distributed file system such as Hadoop to maintain the Big Data.
Fig. 2. Data growth through technological evolution
OnLine Transaction Processing (OLTP) systems or operational systems are
relation database management systems [4], process a huge number and short on-line
transactions. It maintain data consistency with low redundancy real time data
processing or data streaming [5].
OnLine Analytical Processing (OLAP) systems follows write once and read
more paradigm. It uses very complex queries [4] with many aggregative functions in
the query. It is used in data mining and analytics. Data Cube and OLAP are two
popular models in early analytics [6]. Generally data processing done in long running
data batch jobs, it is called as batch processing [5] Jobs.
Brief history of data evolution with milestones [7]: Megabyte to Gigabyte
(1970-80), database machines generate data in GBs. Gigabytes to Terabytes (1980-
‘00), shared nothing parallel databases to generate data in TBs. Terabytes to petabytes
(2000-10), GFS and Map Reduce to generate data in PBs. Petabytes to Exabytes
(2010 +), IoT, Cloud, Big Data enablers, etc., to generate data in EBs
2.5. Evolution of data warehouse [4]
In the initial days, OLTP systems were used for analytics to get the data insights,
though it is advantages to get insight of real time data but analytical functions with
multiple joins on very large number of records is a performance hit. Having both
operational and analytical functions working on the same data base is a challenge as
usability and performance considerations. In order to make quick decisions from the
data, organizations started adapting Decision Support System (DSS) with help of
tools, algorithms and techniques to get data insights. The data aggregation is done in
the analytical models of the data warehouse and ETL (Extract Transform Load) has
been developed to extract the data, transform the data and load them into the data
warehouse form the operational databases.
Data WareHouse (DWH) [4] is a decision support system built as a data driven
system, enables management to take decisions. Usually it focusses on single subject
of the data sometimes it termed as Data mart.
Unauthenticated
Download Date | 7/5/17 9:02 AM

7
Enterprise Data Warehouse (EDW) [4] is a system based on OLAP model,
intended for analyzing the data and generating the insights out of the entire enterprise
data as form of Reports. EDW will focus on all the subject/ functional areas of the
entire business and define the business rules to derive the insights of the data across
the functions, such as Customer, Product, Sales, Services, Finance, Marketing,
Manufacturing, etc.
2.6. Evolution could computing
Storing the data into the database servers, managing the data, maintaining data centres
and the underlying infrastructure is cumbersome for non IT organizations.
Organizations have to invest a lot on the infrastructure and intellectual property for
maintaining the same. Many of the organizations would like to focus on their core
business rather than maintaining these data centres and required infrastructure, as
they focus to gain profits by just focusing on their business strategies. The idea of
providing all these services for managing infrastructure, platforms and software have
been evolved recently. To manage these services with lots of elasticity and on demand
based services, a provider requires to maintain the infrastructure with distributed
computing power.
Grid Data Computing (GDC) [8] is achieved by set of computers and resources
work for some common requirement [9]. It works in basis of distributed architecture
and non-interactive workloads.
Distributed Data Computing (DDC) is something that usually manage a pool of
computers to work together in by message passing technique to achieve the required
goal. Parallel execution of tasks and consolidation at the end is done before sending
the response.
Cloud Data computing (CDC) [8]: Since data management and data processing
is being increased day by day, it became difficult to individual companies to manage
the data bases and data centres. Based on the demand companies need to increase the
data processing and management capabilities and it is the main factor for evolution
of Cloud computing [9]. The idea of cloud computing is to provide all the required
infrastructure as a service on demand within a short time. This will greatly reduce the
investment on an infrastructure by an individual company. Third party data centres
are the hub for Cloud computing operations such as data storage and solutions
provided to users and enterprises. Cloud computing is a shared infrastructure and
services among the cloud users, hence it will be most economical way for data storage
and processing. Undoubtedly, Cloud computing is one of the technology swifts in
this century [5]. Three Characteristics of cloud computing are: 1) Pay per usage;
2) Flexible, and 3) Infrastructure Management [5].
To store, manage and process the data on the cloud it uses terrestrial servers
across the internet. Cloud is a distributed parallel computing mechanism works in-
line with Grid computing technology. These servers collectively process the data.
Cloud computing is synonymous to Service Oriented Computing (SOC). Also, all the
resources such as software, platform and hardware are delivered as a service. It is
provisioned over the internet or other private networks. Three most popular cloud
Unauthenticated
Download Date | 7/5/17 9:02 AM

8
computing components are Software as a Service (SaaS) [5], Platform as a Service
(PaaS) [5], Infrastructure as a Service (IaaS) [5].
Cloud deployment models: Cloud services can be deployed in the four models
such as: Private Clouds [5] – a model where cloud infrastructure is being operated
and managed solely for a single organization. Public Clouds [5] – a model where the
services are available on the internet and is open for public; concern is the security in
the public clouds as it is a shared pool for public. Hybrid Cloud [5] – a combination
of both private and public cloud. Community Clouds – a model for like-minded
organizations with the common goals share the infrastructure, services, etc., among
them.
2.7. Modeling towards Big Data
There are many definitions for Big Data, we may refer to huge data sets where the
size of this referred data is not in a position to manage by any traditional data base
management systems for capturing, storing, analysing and managing the datasets. We
come across data in every possible form, whether it is through social media sites or
sensor networks or digital images or any videos or cell phone GPS signals or purchase
transaction records or web logs or individual medical records or any archives or may
be military surveillance or e-commerce data, complex scientific research data or any
other form. Altogether it amounts to over some large data! This data is what we may
call as…BIG DATA!
3. Big Data – A deep dive
As discussed in the previous section, Big Data is something like a set of huge data
sets which are complex and requires tedious jobs to capture, store, process and
analyse them. The definition of “BIG DATA” may vary from organization to
organization, person to person depending upon their use cases and their value
generation from their data and data characteristics such as data size, capacity,
competence of human resource, techniques used for analysis and, etc. For example,
some organization, managing few GB of data may be a cumbersome job where as for
others it may be some terabytes. Big Data may be referred to data which is being
generated in a very large quantity / volume at a high velocity/ rate in many different
formats of data. As we store the data, we have to plan to analyse the data and get the
insights of it and take some decision based on the insights what we get from the data.
However, as a traditional practice only some of the sample data taken for analysis
instead of taking complete data due to many technical challenges to handle complete
data set. Data was sampled and analysis was done on the sample data for decision
making. However, with the help of Big Data and associated technologies and
frameworks such as Hadoop, we would be able to process and analyse complete data
set. So, we can achieve very accurate results from the complete data set as it is not
biased for decision making.
Data is rapidly growing and to quantify the data units also increased day by day.
As mentioned in Table 1, the different units and its decimal and binary equivalent
Unauthenticated
Download Date | 7/5/17 9:02 AM

9
values shown. Yottabyte is the largest unit with a decimal value of 10 power 24 and
binary value of 2 power of 80.
Table 1. Data volume
Name
Decimal
value
Binary
value
Name
Decimal
value
Binary
value
Kilobytes (KB)
103
210
Petabyte (PB)
1015
250
Megabytes (MB)
106
220
Exabyte (EB)
1018
260
Gigabytes (GB)
109
230
Zettabyte (ZB)
1021
270
Terabytes (TB)
1012
240
Yottabyte (YB)
1024
280
In a study IDC digital universe has predicted that the growth of the data would
go up to 40K Exabytes that is the 50 fold growth of the data by 2020 [10]
Digital data generation starts from mainframe computers, the data generation
speed and volume are very limited. There were few thousands of users and few
hundreds of applications at that time, later client server technology comes into
existence and data speed and volume of the data also increased. With client server
technology, thousands of applications come into existence and millions of users are
using these applications. With an internet and web technologies multiples of ten
thousand of applications are being used by hundreds of millions of users. With
internet and mobile communication come into lime light, lots of data generated with
mobile technologies, social media, sensors, clouds, etc., coming into this generation
we are rapidly generating very huge data sets in different formats.
Fig. 3. Data transformation from main frames to Big Data and data generation every 60 seconds
As mentioned in Fig. 3, 204 million emails are sent every 60 seconds, 2.78 million
of status updates are happening every minute [11]. Consider 90% of the data in the
world now is generated in the last two years. This enormous data has left the
organization to grapple how to make use of this data.
3.1. Need for Big Data
Big Data is becoming very important in our day to day life. It is also becoming really
critical to us as it is not just emerging as one of the most powerful technologies in
this era, also it is helping us to take day to day decisions. Big Data enables to analyse
Unauthenticated
Download Date | 7/5/17 9:02 AM

10
the complete data set instead of some sample data, hence there is non-biased
outcomes. Social media networking enables analytics in a very cost effective manner
and quick feedback and reviews from customers. It is possible to build a product
based on customer preferences with the help of Big Data. In the healthcare centres,
doctor can provide the solution referring to the previous history of the patients. Big
Data getting Normalized [12]. Here Normalization is referred as a process that
enables to generate ideas and actions from the Big Data and it is taken as just a normal
process as day to day life. This becomes a standard practice and part of our lives.
This may be used to disclose the habits of one’s eating, acting and thinking based in
a situation and provides appropriate suggestions.
Fig. 4. Big Data spread
Big Data Presence: As shown in Fig. 4, Gartner has done a survey [11] and
described the presence of Big Data across the world. North America region top
among the other regions 37.8% are adapted Big Data and 18.5% are ready to adapt
Big Data. Latin America being the down side among the others adapted 17.8% and
planned to adapt 11.1% soon.
3.2. Characteristics of Big Data
As depicted in the Fig. 5, Big Data has defined by the below seven characteristics
[13, 14] such as Volume, Velocity, Variety, Veracity, Validity, Volatility and Value.
Volume [1, 15, 16]: Big Data means to someone as enormous volume of data.
As we have discussed Big Data deals very large datasets, it is referred to large
volume. For an instance, 15 terabytes of posts on Facebook could mean Big Data!
Variety refers to data in different forms or types [1, 15, 16]. Data such as emails,
videos, pdf’s, logs, images, excel sheets, etc., are different varieties of data. Any data
can be organized into one of the below mentioned three types [5] such as structured
data, semi-structured data and unstructured data.
Unauthenticated
Download Date | 7/5/17 9:02 AM

11
Fig. 5. Characteristics of Big Data
Velocity is pace at which data generated and flowing from the data sources
[1, 15, 16] like business process, machine generated data and human interactions with
computer, etc. In simple data generated by machines, processes, human interactions,
sensor generated data are generated at a high velocity. Velocity means the rate at
which data is flowing in to the systems. So, Big Data requires fast processing to deal
with the velocity of the data. Time for processing this data plays a very crucial role
as we need to take right decision at the right time in any of the organizations.
Veracity refers to the biased such as partial, falsity such as inaccuracy, noisy
[1, 16], such as unclear and abnormalities in data. It is very challenging job to deal
with veracity in data for doing analysis. It requires to clean and process the data make
the data certainty and truth and consistent. Veracity refers to the uncertainty of the
data, that is the data can sometimes get messed up and it is very difficult to interpret
and trust the data. Due to uncertainty of data, trust worthiness comes in to play and it
is not considered for decision making.
Validity is like veracity, where data has anomalies, other issue of the Big Data
is validity of the data at that moment. This means the data what we use is correct and
accurate for the intended use at the given point. Hence the valid data is key for making
the right decisions.
Volatility in Big Data, may be referred to how long is the data valid and how
long must it be stored in our systems. Organizations would need to plan and
determine at what point the data is no longer relevant for the data analysis.
Value refers to the processing of the data that produced out after analysis [1].
The value of data is not for just one time use it can also be reused for any future
purpose analysis by combining with other data sets.
3.3. Big Data databases
Today lots of data generated with personal information, social graphs, geographic
location, user-generated data though social and mobile media communications and
tools such as Twitter, Facebook, WhatsApp and Google+, etc. Data is exponentially
increasing on day by day with all social networking data, sensor data, multimedia
data and enterprise data. To avail the benefit of the Big Data properly, it is required
to store and process huge number of datasets. These kinds of datasets can’t be handled
by traditional SQL databases, as these were never designed to cater these kinds data.
Unauthenticated
Download Date | 7/5/17 9:02 AM

12
Hence the new generation of No SQL databases are evolved, which can handle these
huge datasets, high velocity data and wide varieties properly.
Data lakes are repositories, where set of multiple data storage instances from
several data assets in an organization [5] are stored. The major role of a data lake is
to ensure all the data in a crude form from all the data sources of an organization.
Simply it is a dump of all the data into one cluster. Data lakes can store any data
types. It is a schema less approach and it tuned for data retrieval in a faster way. Since
load first paradigm [17] enable an enterprise to consume much data from multiple
sources and technologies, this is being enabled through data lakes.
No SQL to be read as Not only SQL [18], is a non-relational database
management systems [19]. These databases are different from traditional relational
database management systems in some ways. No SQL is designed for storing huge
data sets in distributed environment. These type of data bases may not require to
have fixed schema, it avoid join operations and typically it can scale out approach.
Nowadays lots of unstructured data generated and we need to have some way to store
them into databases. No SQL is the solution for storing such types of data. It is clear
that, in the future NoSQL databases going to be more widely used to handle all such
data as schema-less database would be more pronounced [19]. Some of the prominent
and widely used NoSQL Databases are MongoDB, Redis, DataStax, MarkLogic and
DynamoDB, etc.
The comparison of RDBMS with NoSQL database is shown in Table 2.
Advantages and flexibilities of No SQL databases are mentioned in the below table.
Table 2. Comparison between RDBMS and No SQL
RDBMS
No SQL
Structured and organized data: Data stored
in columns and rows
Stands for Not Only SQL, it can store any
structured, semi structured and unstructured data
Structured Query language (SQL)
No declarative query language
Data and its relationships are stored in
separate tables. Pre-defined schema
No predefined schema
Data Manipulation Language, Data
Definition Language
Key-Value pair storage, Column Store, Document
Store, Graph databases
Tight Consistency
Eventual consistency rather ACID property
BASE Transaction
Unstructured and unpredictable data
Based on CAP Theorem
Prioritizes high performance, high availability and
scalability
3.3.1. CAP Theorem
No SQL database is a distributed architecture follows CAP theorem, built on three
characteristics depicted in Fig. 6, such as:
Consistency: This means that the data in the database remains consistent after
the execution of an operation.
Availability: This means that the system is always available, which means no
downtime.
Partition Tolerance: This means that the system continues to function even the
communication among the servers is unreliable.
Unauthenticated
Download Date | 7/5/17 9:02 AM

13
Fig. 6. CAP Theorem
In Practical it is impossible to fulfil all three requirements stated above. CAP
provides the basic characteristics for a distributed system to follow two of the three
characteristics. Therefore, all the current NoSQL database follow the different
combinations of the C, A, P from the CAP Theorem. The advantages and
disadvantages of No SQL are shown in the Table 3.
Table 3. No SQL advantages and disadvantages
Pros
Cons
Highly scalable as it Scale out horizontally
No proper standardization
Distributed Computing
Limited query capabilities
Lower cost as it can work on commodity
hardware and scale out approach
Joins are not possible
Flexible Schema adoption for semi-structure data
No complicated Relationships maintained
3.3.2. Categories of No SQL
NoSQL data bases can be categorized into four general types. Each of the types has
its own pros and cons based on their specifications. Based on the requirement need
to choose the best type suits our requirement. We can’t distinguish one is superior to
other type. There are: 1) Key-Value stores; 2) Column-oriented databases;
3) Graph data base; 4) Document oriented data base.
Key-Value store: It is one of the basic types of No SQL data store [5, 7]. It is
designed to manage large datasets. It was built on the specifications provided in
Amazon’s Dynamo[5] white paper. Since it is key value stores, it enables schema-
less data storage capability. In this, data is represented as hash table. The values in
the table are stored as JSON, String or BLOB, etc. The key of the hash table usually
string or any other object such as hashes, list, set, etc. It will have key and value pairs
of the data being stored for each row. Key-Value stores Database mostly follows the
‘Availability’ and ‘Partition’ aspects in CAP theorem. This kind of Key-Values stores
suitable for shopping cart contents, etc. Example of Key-value store Data Base:
Redis, Dynamo [5], Riak, etc.
Column oriented databases primarily work on columns unlike RDBMS
databases [5, 7]. Each column is here treated independently. In this all the data for a
column are stored contiguously. Column specific files for each column is maintained.
Query in column oriented DB is built and work on columns. Since the data with in a
column will have the same type of data it is easy for comparison within a data file. It
Unauthenticated
Download Date | 7/5/17 9:02 AM

14
gives better performance in querying the details it can directly access specific column
data from the column data file. Row stores data in a row format. It is easy to add or
modify a record, however we need to fetch some unnecessary data to read. Row-
stores are used usually see in RDBMS databases. Whereas column stores used for
write once and read many databases. It is suitable for read intensive large data
repositories such as data warehouses. Hence the column aggregation functions such
as count, sum, average, minimum and maximum will have a greater performance
experience with column oriented databases. Majorly these databases are seen in data
warehouses for building business intelligence from Customer Relationship
Management (CRM), Library card catalogues, etc. Example of Column-oriented
databases: BigTable, Cassandra, SimpleDB, etc.
Graph data base is a hybrid data base [5], which can be backing unusual types
of data store such as graph in data stores [7]. This structure follows set of ordered
pairs called as edge and arcs for each entity. This entity is called as vertices or node.
It has a capability to represent any kind of data elegantly with high accessibility. It is
a collection of nodes and edges; each entity is represented by a node or vertices and
every connection is relationship represented by an edge in the graph between two
nodes or vertices. It is uniquely identified by each node and its edge definition. All
the adjacent nodes will be known by the neighbour nodes. It uses indexing for look
ups. Though the number of nodes increases no of hops remains same. It is an easier
way to represent any hierarchal data in a graphical representation and easy to denote
and fetch the hierarchal information from the graph databases. Example of Graph
databases: OrientDB, Neo4J, Titan, etc.
Document oriented data base is a collection of documents [5]. Data model is
represented in the document itself. It is a collection of key value pairs in a document.
Key allows to access the value from the document. It follows flexible schema and
ease of change in schema in document oriented data bases [7]. In order to group the
different kinds of data, documents are stored into collections. Document maintain
data in a different key value pairs or key and array pairs. One of the important features
of the document oriented database is, it offers nested documents as well, which means
essentially, we can have a document as a value for one key in the document. This
nested document is generally used for representing the hierarchal data or complex
data sets. The comparison between document model and RDBMS is shown in
Table 4. Collection in the document model represents table. Each document is like a
row of a table. Column and its values are represented by key value pairs. Generally,
document model will not support joins.
Table 4. Comparison of RDBMS with document model
RDBMS
Document Model
Tables
Collections
Rows
Documents
Columns
Key/value pairs
Joins
Not available
Examples of Document Oriented databases are MongoDB, CouchDB, etc.
Unauthenticated
Download Date | 7/5/17 9:02 AM

15
3.4. Big Data usage across industries
Most of the organization across the sectors are exploiting the Big Data benefits. It is
aggressively used in certain sectors and they are getting benefited out of Big Data
and analytics. Majorly banking, government, media and communication,
manufacturing, retailers, e-commerce and social media sectors are largely using the
Big Data. Now health care providers, Insurance companies and transportation
companies are emerging in the Big Data analytics usage.
Fig. 7. Big Data – Heat map
Fig. 7, compares the Big Data adoption by different industries [20]. Banking,
securities, communication and media, manufacturing and natural resources are the
majorly adopting Big Data analytics. Big Data usage is very hot in these industries.
There are many use cases exists in these industries and they are already enjoying the
benefits of Big Data. Industries such as government sectors, retail, transportation and
utilities are also using Big Data heavily. Education and wholesale traders’ industries
are having low presence in the Big Data area.
3.5. Analytical data
Organizations using analytics for taking proper decisions to improving their business.
Organizations can analyse the data and identify customer buying patterns, customer
churn outs, profitability and performance characteristics of their suppliers such as
over time, better supply chain management to classify them. Data analytics enables
to identify the failure patterns of a product based on product behaviour data by
scrutinizing the product behaviour for some period of time. All these data to be stored
in an enterprise data warehouse with a support of data marts and with the support of
aggregation functions and algorithms. Enterprise data warehouse will have a large
number of fact tables surrounded by some of the key dimensions such as customer,
supplier, account, location, product and partner, etc., for building their analytics and
business intelligence.
Unauthenticated
Download Date | 7/5/17 9:02 AM

16
4. Big Data analytics
4.1. Data analytics
Data Analytics (DA) [21] is a science, which will examine the data and draws insights
out of the information from the data. Eventually it will enable us to monitor the data
and make us understand what has occurred, why is it occurred, what is going to occur?
and what is to be done to avoid this occurrence?, etc. Hence data analytics is used by
industries to make better business decisions based on the historical data using some
data analytics tools. Data analytics is renowned from data mining by its scope, focus
and the purpose of the data analysis. Data Mining is the process to discover the
patterns in the data and establishing unknown relationships among the data. Data
analytics extrapolates the conclusion based the knowledge of the person or system
with a prior knowledge. Data analytics is used to derive insights of the data starting
from OnLine Analytical Processing (OLAP) to CRM analytics. For an example,
banks will analyse the data of their customers on their usage patterns of money
withdrawals, spending habits, etc., to prevent any fraud in transactions. Another
example on any e-commerce site, they would like to understand the buying patterns
of the consumers in the site and offer recommended products and provide some
discounts to attract the consumers
4.2. History of Big Data analytics
The Big Data concept has been around for some years. Now many of the
organizations understand the importance of Big Data, so they started capturing all the
data that flood into their businesses. These huge data can be applied for analytics and
get significant value from it. Interestingly even before few decades anyone uttered
the term “Big Data,” many organizations used some basic analytics using
spreadsheets, examined the data and uncovered lots of insights and trends from the
data captured in the spreadsheets. Till few years ago companies used to gather
information, run the analytics on the data gathered and reveal the information used
for future decisions. However today organizations are identifying insights for
immediate decisions on the whole data they have been using the Big Data capabilities.
Abilities like quick action, stay agile will boost organizations into a competitive edge.
4.3. Why is Big Data analytics important?
Big Data analytics is the way to help organizations harness their data to identify new
opportunities. That, in turn, the Big Data analytics insights and decision may leads to
smarter moves for business growth, improves day to day business operations,
manages to get higher profits and happier customers and improves the customer
satisfaction. As per Tom Davenport, IIA Director of Research, mentioned in his
research article that many of the companies got value in the following ways with Big
Data analytics as shown in Fig. 8.
Unauthenticated
Download Date | 7/5/17 9:02 AM

17
Fig. 8. Big Data analytics – value
Cost reduction: Hadoop and Cloud based analytics are Big Data analytic
platform and technologies, which can bring in significant cost advantages especially
for storing large amounts of data and process. Also by performing analytics on the
Big Data they can clearly identify the ways for doing efficient business.
Faster: Better decision making: For a faster processing of data and analysis,
new sources of data, technologies like Hadoop and in-memory analytics companies
are able to analyze information real-time so as the decisions are taken at the right time
based on what they’ve analyzed.
New products and services: Through Big Data analytics organizations gain
ability to gauge customer needs and understand the sati factionary requirement. With
this in mind it enables the organization to think what their customer wants. Many
companies are manufacturing new products to meet customers’ needs based on the
analytics.
4.4. Analytics on Big Data [15]
Just having Big Data alone is useless, you get some insights out of them which is
required for taking any business decision to the growth of the enterprise[22]. We
need tools and frameworks to highlight the insights from huge datasets, rapidly
increasing data and a large variety of data of an enterprise. In the subsequent sections
of this document, discussed on the methods and tools used for analysing Big Data
analytics in the industry.
Big Data analytics have been embraced as a disruptive technology that will
reshape business intelligence which is a domain that relies on data analytics to gain
business insights for better decision-making. There is a need of state-of-the-art Big
Data analytics tools which are much innovative, highly scalable and powerful for
analyzing into insights with multidimensional data visualization [5]. There are many
tools built by many people for Big Data visual analytics. Few of them are open source
and some of them are commercial. For an instance IBM’s InfoSphere and BigInsights
[5] are commercial licensed tools for Visual analytics.
4.5. Types of Big Data analytics
There are four types of Big Data analytics: Prescriptive Analytics – This type of
analysis reveals what actions should be taken; this is the most valuable kind of
analysis and usually results in rules and recommendations for next steps. Predictive
Unauthenticated
Download Date | 7/5/17 9:02 AM

18
Analytics – An analysis of likely scenarios of what might happen; this kind of
analytics are usually a predictive forecasting. Diagnostic Analytics – A look at past
performance to determine what happened and why it happened. Descriptive Analytics
– What is happening now based on incoming data. To mine the analytics, you
typically use a real-time dashboard and/or email reports; this is like a report,
dashboards, score cards, etc.
4.6. Techniques and methods for Big Data analysis [23]
Based on statistical and mathematical models, some computerized techniques have
been developed, which can be used to analyse datasets and provide some insights of
the data as a result. There are numerous techniques available in Big Data analytics.
All these methods and techniques are being enhanced day by day by the researchers.
Review of few techniques and methods used in Big Data analytics are provided in
Table 5
Table 5. Big Data analytics techniques and models
No
Technique/
Method
Details
1
A/B Testing
This is called as split testing or bucket testing. In this technique, a
control object is compared with many test objects for improvements.
This is not being used widely in Big Data
2
Associated Rule
Mining
This technique is used for relationships among the huge datasets. This
would use many algorithms to generate association rules. Ex. Market
basket analysis – it would provide details to a retailer how customers
are buying two or more different products together. Which will define
an association between those two products. This technique used in
data mining
3
Classification
This is a technique to identify the category of the incoming new data
based on the existing set which are already categorized based on
some data points. This is often used as supervisory learning technique
as it uses existing data sets called training dataset. This technique is
also used in data mining
4
Cluster Analysis
This is a technique used to group a set of objects into same group
based on some commonalities to each other and different to other
groups. This technique is used in data mining
5
Crowdsourcing
This is a technique used to generate the ideas/ innovations from a
large group through an online method. It creates a new knowledge
repository for ideas
6
Data Fusion and
Integration
This is a technique used to generate more efficient and potential
insights of the data using set of models and techniques which
integrate multiple data sets and analyze, rather analyzing them
independently
7
Data Mining
This is major technique used for predictive analytics. This enables to
identity hidden patterns with help of association rule, classification,
regression and cluster analysis on large datasets
8
Ensemble
Learning
It is a type of supervisory learning technique. It uses many predictive
models such as machine learning and statistical approach to gain
better predictions than simple constituent model
9
Genetic
Algorithm
This technique is based on the process of Natural evolution
Unauthenticated
Download Date | 7/5/17 9:02 AM

19
Table 5 (c o n ti n u e d)
No
Technique/
Method
Details
10
Machine
Learning
This is a technique used to create artificial intelligence by providing some
knowledge to the system. The knowledge is based on the algorithms that
enable systems to progress with the help of pragmatic data (knowledge).
It is aimed to identify the complex pattern and take appropriative decision
based on the intelligence of the system (knowledge)
11
Natural
Language
Processing
It is a kind of Machine Learning process. NLP uses set of techniques as
defined in Machine learning. Ex Sentiment analysis on a social media to
find out, how customers are reacting to campaign
12
Neural
Networks
It is computational techniques, which was mimic of structure of biological
neural networks. It is suitable for nonlinear pattern recognition. It uses
both supervised learning and unsupervised learning
13
Network
Analysis
This is a technique used in a network or a graph to illustrate the
associations among the discrete nodes
14
Predictive
Modeling
This is a model built with the help of some of the statistical and
mathematical techniques to predict a best possible outcome
15
Regression
This is a statistical technique used for prediction which will determine co-
variance between dependent and independent variables
16
Sentiment
Analysis
This technique is a Natural Language Processing which can determine the
information of subject from the source of textual data
17
Spatial
Analysis
This Technique uses set of statistical models which can be used to explore
the geographical data such as Geographical Information Systems (GIS)
which used to capture the longitude and latitude, i.e., exact location
18
Statistics
Statistics is a science of collecting data in the form of surveys or
experiments etc., organizing the data in any sorting order or lexical order
etc., with an art of interpretation of the data in a form of hypothesis
19
Supervised
Learning
This is a technique which adapts some of the machine learning techniques
which can deduce relationships based on prior knowledge (training data)
20
Simulation
This model is used to mimic the behavior of some complex systems to
predict and plan the outcome and measure the results
21
Unsupervised
Learning
It is a technique uses some of the machine learning models which can
identify the hidden patterns from the data without any prior knowledge.
An example of Unsupervised learning is Cluster Analysis
22
Visualization
This technique used to provide a gist of the information in very simple
way in terms of a graph, diagram, image or any visual representation to
simplify the understanding
Since Big Data is a combination of structured, semi-structured and unstructured
data sets, it is important to think through various analytical data models for analyzing
them. Relational data (structured) can be analyzed with Structured Analytics which
is being used commonly to analyse OLAP data. Text analytics are the analytics used
for social media analytics, web analytics. Social media analytics is majorly used for
sentiment analysis. There are several models available for performing sentiment
analysis. There are three types of sentiments [24], positive, negative and neutral,
which can bring in a potential business values. Social media data fed by consumer
for a product of an organization help the organization to understand the customer
sentiments and plan their strategies to improve their sales and product quality. Visual
analytics used for analyzing multimedia data such as visual shapes generated by geo
special systems, etc., are briefly discussed in the next sections.
Unauthenticated
Download Date | 7/5/17 9:02 AM

20
4.7. Text analytics
There is an increasing generation of text data in the form of unstructured data such as
social media messages, web logs, different forums sharing the feedbacks about some
product and email communication, etc. These scenarios have evolved from the
traditional structured data usage using RDBMS and Data Warehouses [25] to cater
semi structured and unstructured structured data. With digital transformation, IOT,
etc., over 90% of data generated is semi-structured or unstructured. So, it is important
to use the analytics on these varieties of data. Text analytics is process having
multiple steps starting from text data identification to visualization. As shown in
Fig. 9, 10 steps involved in text mining, same is described in the next paragraph.
Fig. 9. Text analytics in Big Data
Ten prominent steps involving the Text Analytics (TA) are: 1. Collecting Text
Data from the different Sources. 2. Extracting the concepts, entities, events and
relations. 3. Creation of taxonomies out of them. 4. Search accessing, web crawling,
indexing, de-duplicating from the data clustered and categorized. 5. Analyze the file
format using nature language semantics. 6. Build ER modeling. 7. Link the analysis
among them. 8. Ability to identify sentiments out of this. 9. Document the
summarized information. 10. Visualize them using the tolls to have quick insights
Text Mining (TM) is process to analyze text data to derive possible insights from
text data content such as emails, text documents, text communication in the social
networking such as postings, feedbacks, etc. Mining unstructured data with Natural
Language Processing (NLP), statistical modelling and machine learning techniques
is a challenging task, because of natural language text is often inconsistent. Text data
may contain some ambiguities due to inconsistent syntax and semantics, including
slang, specific business languages, language by different age groups, etc.
4.8. Visual analytics
Visual analytics applications are mostly having huge and high dimensional data sets
[26] such as climate research, geo special research and financial market research, etc.
Recent developments in visual shape analytics has been a considerable success which
requires certain algorithms for analysis for navigating and visualizing that are capable
of interactive performance.
Unauthenticated
Download Date | 7/5/17 9:02 AM

21
The main goal of visual analytic’s [27] is to provide insight from very huge
datasets such as scientific researches, forensic data, academic records, any business
data, HTML/ XML files, Web pages, metadata of any visual database and source
code, etc., which are stored in many heterogeneous data stores like NoSQL databases.
To bring in the statistical analytics from these data, it is required to iteratively collect
and pre-process data [27], however to enable the decision making from these data
one has to acquire knowledge and get a perception out of the data by exploring
different knowledge representations of the data along with human capacities. Hence,
the human factor is a prime factor in the field of visual analytics. The Human
Computer Interaction (HCI) [28] is a crucial component that supports knowledge
discovery from the visual analytics. HCI is all about the building knowledge of
interface between human and computer. The other component in this field is Semi-
autonomous system [28], in this model enables to run the system independently for
under assured conditions, however this can’t take end to end decisions.
4.9. Big Data use cases
There are many tools, techniques, solutions and frameworks are available to
implement Big Data and are developed with the references to heterogeneous
architectures [29]. Social networking companies such as Linked In, Facebook,
Twitter are widely using the Big Data use cases [29].
Table 6. Big Data use cases – industry wide
No
Industry
Use cases
1
Health Care
 Trends of QoS in relevance with health habits[31]
 Clinical Decision support system
 Remote Patient monitoring
 Patient Profiling
 Comparative effective research
 Disease prediction[32]
2
Marketing
 Cross Selling
 Location Based Marketing
 Analyzing instore behavior
 Sentiment Analytics
3
Manufacturing
 Sensor based operations
 Supply chain and inventory management
 Shorten driven to value cycle
4
Public Sector/ Govt
Sector
 Identifying the basic needs from the public[32]
 Traffic decongestion
 Civic compliance to reduce Noise, Air and Water pollution
5
Banking & Insurance
 Fraud detection
 Customer Predictions
 Risk analysis
6
Telecom
 Geo Targeted Ads
 Emergency response
 Remote monitoring of personal things
 Urban planning
7
Oil & Gas
 Remote drilling
 Oil Prediction
 Digital Oil field
Unauthenticated
Download Date | 7/5/17 9:02 AM

22
Banking sectors also use many of the use cases, for an instance a bank requires
to predict its customer behaviour for successful banking business. A framework
called iCARE [30] made an attempt to analyse the unified customer behaviour and
their transactions to predict the customer behaviour and giving insights for building
new strategies to their business growth. Here is given some of the Big Data use cases
used in important industries as detailed in Table 6.
5. Big Data solution
5.1. Apache hadoop – history, architecture and eco system
Hadoop [33] was developed by Dough Cutting along with Mike Cafarella in 2005
[34]. Based on Google’s white paper Google’s distributed File System (GFS),
Hadoop [14] is a distributed and parallel computing paradigm built based on Google’s
MapReduce paradigm [35]. It is built to process a very huge data sets with in a
clustered environment on scale out approach, which means essentially, we can add
thousands of machines on to the hadoop cluster, so each offering local computation
and storage and can process the data in a parallel.
The Apache Hadoop framework is modules as a master and slave kind of
architecture. It contains the following components: 1) Hadoop Distributed File
System (HDFS) – this module is the one which has actually store the data in a
distrusted fashion; it enables the users by providing access to the data with a high-
throughput. 2) Processing the data using (YARN) [35, 36]: Yet another resource
negotiator is module provides resource management in the cluster. 3) Hadoop Map
Reduce – this module works on top of YARN for processing the large data sets in a
parallel processing mode.
Hadoop Distributed File System (HDFS) is the file system in Hadoop
framework. It is designed in a distributed approach and to run on commodity
hardware [37]. HDFS is highly fault tolerant on low cost hardware. It can store a large
amount of data and provides easy access to the users. Master Slave architecture [38]
is followed in HDFS. Name node is a master server manages the name spaces in the
file system as a meta data [34]. Data Node is a slave system which manages the data
storage of the system [34]. In a cluster, there will be multiple data nodes which can
be used for data operations. Block Size is the minimum amount of data is being read
or write by HDFS file system at once. The size of the block is generally 64 MB [34].
MapReduce [39] is a framework used for processing the batch processing jobs
in the HDFS work on parallel processing mode in the cluster [40] and it is
programming model [38]. The program breaks the execution into two different parts
or phrases such as Map phase and Reducer Phase [34]. YARN - Yet Another Resource
Negotiator is cluster management system in the Apache Hadoop later versions [39]
Some related projects/tools in Hadoop ecosystem are depicted in the following
Fig. 10.
Unauthenticated
Download Date | 7/5/17 9:02 AM

23
Fig. 10. Hadoop technology stack
Apache Hive [39] is a data warehousing package[5] that is built on top of
Hadoop [41] for creating database, tables/views, etc. This is mainly used for data
query and analytics [34]. It is very similar to SQL, which is called HiveQL [42, 43].
Apache Pig [39] an open source high level data flow system for analysing huge
data sets due to amendable structure it enables greater parallelism [44]. This tool is a
high level query and analysis tool built to overcome the obstacles of writing map
reduce code for non-java programmers Pig was created [5]. The language is called
Pig Latin [34]. This will be compiled into map reduce jobs by Pig compiler that will
run on Hadoop [45].
Apache Flume is a data acquisition tool for data integration [46], which is
distributed, highly reliable and with simple configurations and primarily for
streaming data processing such as log data from various web servers to HDFS [34].
Apache Sqoop is a data acquisition tool designed for efficiently transferring the
bulk data between hadoop and structured data stores such as relational database [47].
Apache Spark [36] is a lightning-fast cluster computing framework designed for
fast computation. It can run on standalone, hadoop, mesos or even cloud [48]. Spark
is lightning faster because of the fact it is built in-memory computing framework to
address one of the original hadoops disadvantage iterative operations [49]. In memory
cluster processing and cluster computing is the prime feature of Spark [50], hence it
can have greater processing speed. It covers batch, streaming, interactive and iterative
workloads [51]. Resilient Distributed Datasets (RDDs) is core data units in Spark
[50]. Spark is becoming popular as it has features such as speed, multi-language
support and analytics support.
Apache Zookeeper is a software project from Apache, providing an open source
distributed configuration service, synchronization service and naming registry for
large distributed systems [52]. It will manage and co-ordinate clusters (like Hbase,
Hadoop, Solr, etc.).
As mentioned, there are many advantages of Hadoop usage for Big Data
challenges and few of them are given in brief:
1. Distributed data storage and processing framework work with computation
done at local to data which can prevents the network overload.
2. Linear scale out compared to RDBMS or any other solution.
3. Fault tolerant and high availability.
4. Ability to process in parallel and very rapidly of large data sets.
5. Has ability to process wide varieties of data and stream of data.
6. It best suits for data ware housing for loaded data into Hadoop [53].
Unauthenticated
Download Date | 7/5/17 9:02 AM

24
Some of the disadvantages of distributed computing using Hadoop:
1. Map reduce programming model is very restrictive due to lack of central data
can be preventive as it is not meant for iterative and recursion.
2. Joins of multiple datasets are tricky and slow hence there is often entire
dataset gets copied in the process, which means essentially de normalize the data for
better performance.
3. It will not work well with small and many data sets.
6. Conclusions and further scope
This review article we have discussed overview of data transformed to Big Data,
challenges and issues in Big Data processing and how to harness the data using the
Big Data analytics and models. We have featured some of the solutions for Big Data
challenges with technology advancements and their implementations in today’s
world, being used by industries to provide state of art and near real time data analytics
solution with high precision enhancing the ability to understand the end user
experience and provide better business models.
However, there are some open challenges with these Big Data solutions such as:
1) fast changing system configuration requirements due to highly dynamic workload
constraints, varying innovation cycles of system hardware components, 2) low
latency querying, 3) transactional data handling, 4) data normalization is not feasible
etc. which are not addressed in the current solution. There is an interesting proposal
by H.-K. Lin and others, an architectural system called composable systems which
can potentially address this challenge by rack scale architecture [53]. Secondly, there
is a need of next gen Big Data solutions which could take advantages for transactional
management in a scaling out model like other, not only SQL to bridge the gap. There
are experiments in the similar lines done by D. Plase and team in order to gain more
detailed experience with compact data formats [54] and an interesting architecture
called Splice Machine [55]. Splice Machine is coming up with a hybrid approach to
take an advantage to support greater concurrency even for small data writes and reads
on a distributed scaled out technologies. Apart from this, other major challenge is
analysing Big Data at real time. The current technologies and frameworks are not in
a position to handle all the real-time scenarios to analyse the data at real time. There
is an attempt made by K. Wang and others for real time analytics using hybrid-stream
a Big Data analytics model [56]. There is an interesting novel proposal for
normalizing the data in Big Data approach by Go lo v and R ön nb äc k [57] using
Anchor modelling technique. This seems to have efficient way to store and process
the data. There is a potential scope for research in the above-mentioned challenges.
Ref er en c es
1. D e mc h e n k o, Y., C. D. L a a t, P. M e m b re y. Defining Architecture Components of the Big
Data Ecosystem. – In: Proc. of International Conference Collaboration Technologies and
Systems (CTS’14), Vol. 14, 2014, pp. 104-112.
Unauthenticated
Download Date | 7/5/17 9:02 AM

25
2. S l av a k i s, K., G. B. G i a n na k i s, G. M at e o s. Modeling and Optimization for Big Data
Analytics: (Statistical) Learning Tools for Our Era of Data Deluge. – IEEE Signal Processing
Magazine, Vol. 31, 2014, pp. 18-31.
3. S h er m a n, R. Chapter 1 – The Business Demand for Data, Information, and Analytics. – Business
Intelligence Guidebook, Morgan Kaufmann, Boston, 2015, pp. 3-19.
4. L i ns t e d t, D., M. O l s c h i m k e. Chapter 1 – Introduction to Data Warehousing – In Data Vault
2.0, Morgan Kaufmann, Boston, 2016, pp. 1-15.
5. S h a r m a, S. Expanded Cloud Plumes Hiding Big Data Ecosystem. – Future Generation Computer
Systems, Vol. 59, 2016, pp. 63-92.
6. C o he n, J., B. D o l an, M. D u nl a p, J. M. H e l le r s t ei n, C. W e l to n. MAD Skills: New
Analysis Practices for Big Data. – Proc. VLDB Endow, Vol. 2, 2009, pp. 1481-1492.
7. H u, H., Y. Wen, T. S. Ch u a, X. L i. Toward Scalable Systems for Big Data Analytics: A
Technology Tutorial. – IEEE Access, Vol. 2, 2014, pp. 652-687.
8. M y er s o n, J. M. Cloud Computing Versus Grid Computing. 3 March 2009.
http://www.ibm.com/developerworks/library/wa-cloudgrid/
9. A l kh a n a k, E. N., S. P. Lee, R. R e z a ei, R. M. P a r i z i. Cost Optimization Approaches for
Scientific Workflow Scheduling in Cloud and Grid Computing: A Review, Classifications, and
Open Issues. – Journal of Systems and Software, Vol. 113, 2016, pp. 1-26.
10. The Digital Universe of Opportunities: Rich Data Increasing Value of the Internet of Things. – EMC
Digital Universe with Research & Analysis by IDC.
http://www.emc.com/leadership/digital-universe/2014iview/executive-summary.htm
11. Kim, L. Here’s What Happens in 60 Seconds on the Internet. 11 December 2015.
http://smallbiztrends.com/2015/12/60-seconds-on-the-internet.html
12. K a rt, N. H. L., F. B u yt e n d ij k. Survey Analysis: Big Data Adoption in 2013 Shows Substance
behind the Hype. – Gartner’s 2013 Big Data Study, 2013.
13. Contributors, W. Big Data. 12 March 2016. UTC.
https://en.wikipedia.org/w/index.php?title=Big_data&oldid=709642525
14. I s hw a r a pp a, J. A n u r ad h a. A Brief Introduction on Big Data 5Vs Characteristics and
Hadoop Technology. – Procedia Computer Science, Vol. 48, 2015, pp. 319-324.
15. W a ts o n, H. J. Tutorial: Big Data Analytics: Concepts, Technology, and Applications. –
Association for Informaiton Systems, Vol. 34, 2014, pp. 5-16.
16. S w an, M. Philosophy of Big Data: Expanding the Human-Data Relation with Big Data Science
Services. – In: Proc. of First International IEEE Conference of Big Data Computing Service
and Applications (BigDataService’2015), 2015, pp. 468-477.
17. F a ri d, M., A. R o a t i s, I. F. I l y as, H.-F. H o f f ma n n, X. Chu. CLAMS: Bringing Quality
to Data Lakes. – In: Proc. of 2016 International Conference on Management of Data, San
Francisco, California, USA, 2016, pp. 2089-2092.
18. D o n K o g a n. Top 8 Bigdata Trends 2016. – White Paper, 2016.
19. R i th, J., P. S. L e h ma y r, K. M e y e r-We g e n e r. Speaking in Tongues: SQL Access to NoSQL
Systems. – In: Proc. of 29th Annual ACM Symposium on Applied Computing, Gyeongju,
Republic of Korea, 2014, pp. 855-857.
20. G a i t h o, M. How Applications of Big Data Drive Industries. – Simplylearn.
http://www.simplilearn.com/big-data-applications-in-industries-article
21. S h er m a n, R. Chapter 15. Advanced Analytics. – In: Business Intelligence Guidebook. Boston,
Morgan Kaufmann, 2015, pp. 375-402.
22. G an d o m i, A., M. H a i d e r. Beyond the Hype: Big Data Concepts, Methods, and Analytics. –
International Journal of Information Management, Vol. 35, 2015, pp. 137-144.
23. M a ny i k a, M. C. J., B. B r o w n, J. B u g hi n, R. D o bb s, C. R o xb u r g h, A. H. B ye r s. Big
Data: The Next Frontier for Innovation, Competition, and Productivity. McKinsey Global
Institute, June 2011.
24. V a tr a p u, R., R. R. Mu k k a ma l a, A. H u s s a in, B. F l e s c h. Social Set Analysis: A Set
Theoretical Approach to Big Data Analytics. – IEEE Access, Vol. 4, 2016, pp. 2542-2571.
25. I t to o, A., L. M. N g uy e n, A. V a n D e n B o s c h. Text Analytics in Industry: Challenges,
Desiderata and Trends. – Computers in Industry, Vol. 78, 2016, pp. 96-107.
26. H e rm a n n, M., R. K l e i n. A Visual Analytics Perspective on Shape Analysis: State of the Art
and Future Prospects. – Computers & Graphics, Vol. 53, Part A, 2015, pp. 63-71.
Unauthenticated
Download Date | 7/5/17 9:02 AM

26
27. G o nz á l e z-To r r e s, A., F. J. G a r c ía-P e ñ a l vo, R. T h e r ón-S á n c he z, R. C o l o mo-
P a l a c i o s. Knowledge Discovery in Software Teams by Means of Evolutionary Visual
Software Analytics. – Science of Computer Programming, Vol. 121, 2016, pp. 55-74.
28. M a ko n i n, S., D. M cV e i g h, W. S tu e r z li n g e r, K. T ra n, F. P o p o wi c h. Mixed-
Initiative for Big Data: The Intersection of Human + Visual Analytics + Prediction. – In: 2016
49th Hawaii International Conference on System Sciences (HICSS’16), 2016, pp. 1427-1436.
29. P ä äk k ö n en, P., D. P a k k a la. Reference Architecture and Classification of Technologies,
Products and Services for Big Data Systems. – Big DATA Research, Vol. 2, 2015,
pp. 166-186.
30. Sun, N., J. G. M o r r is, J. X u, X. Zhu, M. Xie. iCARE: A Framework for Big Data-Based
Banking Customer Analytics. – IBM Journal of Research and Development, Vol. 58, 2014,
pp. 4:1-4:9.
31. B a ta r s e h, F. A., E. A. L at i f. Assessing the Quality of Service Using Big Data Analytics: With
Application to Healthcare. – Big Data Research, Vol. 4, 2016, pp. 13-24.
32. A r ch e n a a, J., E. A. M. A ni t a. A Survey of Big Data Analytics in Healthcare and Government.
– Procedia Computer Science, Vol. 50, 2015, pp. 408-413.
33. S a ra l a d e vi, B., N. P a z h a n ir a j a, P. V. P a u l, M. S. S. B a s h a, P. D h av a c h el v a n.
Big Data and Hadoop – a Study in Security Perspective. – Procedia Computer Science,
Vol. 50, 2015, pp. 596-601.
34. U z un k a y a, C., T. E n s ar i, Y. K a vu r u c u. Hadoop Ecosystem and Its Analysis on Tweets. –
Procedia – Social and Behavioral Sciences, Vol. 195, 2015, pp. 1890-1897.
35. C a ss a l e s, G. W., A. S. Ch a r ã o, M. K. P i n h ei r o, C. S o u v ey e t, L. A. S t e f fe n e l.
Context-Aware Scheduling for Apache Hadoop over Pervasive Environments. – Procedia
Computer Science, Vol. 52, 2015, pp. 202-209.
36. S h ya m, R., B. H. B. Ga n e s h, S. S. K u ma r, P. P o o r n ac h a n d ra n, K. P. So m a n. Apache
Spark a Big Data Analytics Platform for Smart Grid. – Procedia Technology, Vol. 21, 2015,
pp. 171-178.
37. M a, Y., Y. Z h o u, Y. Y u, C. P en g, Z. W an g, S. D u. A Novel Approach for Improving Security
and Storage Efficiency on HDFS. – Procedia Computer Science, Vol. 52, 2015, pp. 631-635.
38. M a it r e y, S., C. K. Jha. MapReduce: Simplified Data Analysis of Big Data. – Procedia Computer
Science, Vol. 57, 2015, pp. 563-571.
39. L o sh i n, D. Chapter 7. Big Data Tools and Techniques. – In: Big Data Analytics. Boston, Morgan
Kaufmann, 2013, pp. 61-72.
40. Y i ld i z, O., S. I b r a hi m, G. An t o n i u. Enabling Fast Failure Recovery in Shared Hadoop
Clusters: Towards Failure-Aware Scheduling. – Future Generation Computer Systems, 2016.
41. Apache Hive TM.
https://hive.apache.org/
42. C h en n a m se t t y, H., S. C h a l as a n i, D. R i l e y. Predictive Analytics on Electronic Health
Records (EHRs) Using Hadoop and Hive. – In: 2015 IEEE International Conference Electrical,
Computer and Communication Technologies (ICECCT’15), 2015, pp. 1-5.
43. X u, Y., S. Hu. QMapper: A Tool for SQL Optimization on Hive Using Query Rewriting. – In: Proc.
of 22nd International Conference on World Wide Web, Rio De Janeiro, Brazil, ACM, Vol. 1,
2013, pp. 211-212.
44. Apache Pig.
https://pig.apache.org/
45. R a ju r k a r, G. D., R. M. G ou d a r. Notice of Violation of IEEE Publication Principles, A Speedy
Data Uploading Approach for Twitter Trend and Sentiment Analysis Using HADOOP. –
In: International Conference on Computing Communication Control and Automation
(ICCUBEA’15), Vol. 1, 2015, pp. 580-584.
46. Apache Flume.
https://flume.apache.org/
47. Apache Sqoop.
http://sqoop.apache.org/
48. Apache Spark.
http://spark.apache.org/
Unauthenticated
Download Date | 7/5/17 9:02 AM

27
49. L i, H., K. L u, S. M e n g. Bigprovision: A Provisioning Framework for Big Data Analytics. – IEEE
Network, Vol. 29, 2015, pp. 50-56.
50. R e ye s-O r t i z, J. L., L. O n e to, D. A n g ui t a. Big Data Analytics in the Cloud: Spark on
Hadoop vs MPI/OpenMP on Beowulf. – Procedia Computer Science, Vol. 53, 2015,
pp. 121-130.
51. E l ia, D., S. F i o r e, A. D’An c a, C. P a l a zz o, I. F o s te r, D. N. W i l l ia m s. An In-Memory
Based Framework for Scientific Data Analytics. – In: Proc. of ACM International Conference
on Computing Frontiers, 2016, pp. 424-429.
52. Apache ZooKeeper™.
https://zookeeper.apache.org/
53. Lin, H.-K., J. A. H a rd i n g, C.-I. Ch e n. A Hyperconnected Manufacturing Collaboration System
Using the Semantic Web and Hadoop Ecosystem System. – Procedia CIRP, Vol. 52, 2016,
pp. 18-23.
54. P l as e, D., L. N i e d ri t e, R. T a ra n o v s. Accelerating Data Queries on Hadoop Framework by
Using Compact Data Formats. – In: 4th IEEE Workshop on Advances in Information,
Electronic and Electrical Engineering (AIEEE’16), 2016, pp. 1-7.
55. Splice Machine.
http://www.splicemachine.com/product/
56. W a ng, K., J. M i, C. X u, L. Shu, D. J. D e ng. Real-Time Big Data Analytics for Multimedia
Transmission and Storage. – In: IEEE/CIC International Conference on Communications in
China (ICCC’16), 2016, pp. 1-6.
57. G o lo v, N., L. R ö n n b äc k. Big Data Normalization for Massively Parallel Processing
Databases. Computer Standards & Interfaces Available Online, 2017. ISSN 0920-5489.
Unauthenticated
Download Date | 7/5/17 9:02 AM