WEB MINING IN E-COMMERCE

Abstract

Recently, the web is becoming an important part of people’s life. The web is a very good place to run successful businesses. Selling products or services online plays an important role in the success of businesses that have a physical presence, like a re

Full text

959
WEB MINING IN E-COMMERCE
Istrate Mihai
University of Pitești Faculty of Mathematics and Informatics No.1, Stadionului, Țicleni, Gorj, 215600, Romania
mihaifrance@yahoo.com +40 745 775 935
Recently, the web is becoming an important part of people’s life. The web is a very good place to run successful
businesses. Selling products or services online plays an important role in the success of businesses that have a physical
presence, like a retail business. Therefore, it is important to have a successful website to serve as a sales and marketing tool.
One of the effective used technologies for that purpose is data mining. Data mining is the process of extracting interesting
patterns from large databases. Web mining is the usage of data mining techniques to extract interesting information from web
data. This paper presents the three components of web mining: web usage mining, web structure mining and web content
mining and the main data preprocessing tasks for web usage mining.
Keywords: E-Commerce, Data mining, Web mining
JEL M
1. E-Commerce and Retail websites
In e-commerce instead of having your business in a limited physical place and a limited sector of customers who
are usually near to your store or business, you have it in the web. In e-commerce websites you have the ability to
sell, advertise, and introduce different kinds of services and products in the web. E-commerce websites have the
advantage of reaching a large number of customers regardless of distance and time limitations. Furthermore, an
advantage of e-commerce over traditional businesses is the faster speed and the lower expenses for both e-
commerce website owners and customers in completing customers transactions and orders.
Because of the above advantages of e-commerce over traditional businesses, a lot of industries in different fields
such as retailing, banking, medical services, transportation, communication, and education are establishing their
business in the web. But creating a successful online business can be a very difficult and costly task if not taking
into account e-commerce website design principles, web engineering techniques, and what e-commerce is supposed
to do for the online business. Understanding the requirements of both e-commerce website owner and customer is
an important aspect in building a successful e-commerce website. There is a lot of information need to be defined
before starting building the e-commerce website such as identifying business goals and how the website will target
those goals, if the website supposed to attract new customers or increase the sales of current customers, identify if
the proposed website will increase the business overall profit, and identify the most suitable tools and techniques
need to be used/followed in order to target those requirements.
Retail websites aim to inspire, reflect a good image about the business and improve it online. An important factor
in having a successful retail website is to know your competitors. On one hand, by identifying their points of
strongness and trying to get benefit of them by improving those strongness points and adopting powerful strategies.
On the other hand, identifying weakness points of your competitors and avoid them is a good practice in having a
successful retail website.
2.Web mining
The usage of data mining to maintain websites and improve their functionality is an important field of study.
Patterns extracted from applying data mining techniques on web data can be used to maintain websites by
improving their usability through simplifying user navigation and information accessibility and improving the
content and the structure of the website in a way that meets the requirements of both website owner and user which
will consequently increase the overall profit of the business.
Web mining is the use of data mining techniques to extract useful patterns from the web. Those extracted patterns
are used to improve the structure of websites, improve the availability of the information in the websites and the
way those
pieces of information are introduced to the website user, and to improve data retrieval and the quality of automatic
search of information resources available in the web. Web mining can be divided into three major categories: web
usage mining, web content mining, and web structure mining.
2.1 Web Usage Mining
Web usage mining or web log mining is the process of applying data mining techniques to web log data in order to
extract useful information from user access patterns. Web usage mining tries to make sense of the data generated
by the web user’s sessions or behaviors. The web usage data includes data from web server access log, proxy server
logs, browser logs, user profiles, registration data, cookies, and user queries. Web usage mining tries to predict user
behavior while user interacts with the web and learns user navigation patterns. The learned knowledge could then
be used for different applications such as website personalization, business intelligence, usage characterization and
adaptive websites. There are two approaches for web usage mining process:

960
- Mapping the log data into relational tables before an adopted data mining techniques is performed.
- Using the log data directly by utilizing special preprocessing techniques.
The Web usage mining process consists of three phases: data preprocessing, pattern discovery, and pattern analysis.
Pattern discovery is that set of methods, algorithms, and techniques used to extract patterns from web log file.
Several
techniques are used for pattern discovery such as statistical analysis, clustering, classification, and sequential
pattern mining. After patterns are discovered they need to be analyzed in order to determine interesting and
important patterns, besides the removal of redundant patterns. Pattern analysis has several different forms such as
knowledge query mechanism, visualization techniques, and loading usage data into a data cube in order to perform
Online Analytical Processing OLAP operations.
A web server log file records users transactions in the web. Usually, the web log file contains information about the
user IP address, the requested page, time of request, the volume of the requested page, its referrer, and other useful
information. The web log file can have different format, but there is a common log file format that is mostly used.
The common log file has the following format:
remotehost rfc931 authuser [date] "request" status bytes
where remotehost represents remote hostname (or IP number if DNS hostname is not available), rfc931 represents
the remote logname of the user, authuser represents the username as which the user has authenticated himself,
[date] represents dateand time of the request, ”request” represents the request line exactly as it came from the
client, status represents the HTTP status code returned to the client, and finally bytes represents the content-length
of the document transferred. The WWW Consortium (W3C) presented an extended format for web server log file
that is able to record a wide range of data to make an advanced analysis of the web log file. Web log file is the main
source of data analysis in web mining but a lot of preprocessing efforts need to be performed in order to prepare the
web log file to be mined.
2.2 Web Content Mining
Web content mining is mining the data that a web page contains. The contents of most of the web pages are texts,
graphics, tables, data blocks, and data records. A lot of research has been done to cover different web content
mining issues for the purpose of improving the contents of the web pages, improving the way they are introduced to
the website user, improving the quality of search results, and extracting interesting web page contents.
Web content mining is still a large field. It constains:
- structured data extraction;
- sentiment classification, analysis and summarization of consumer reviews;
- information integration and schema matching;
- knowledge synthesis;
- template detection and page segmentation;
A large amount of information on the Web is contained in regularly structured data objects which are data records
retrieved from databases. Such Web data records are important because they often present the essential information
of their host pages, lists of products and services.
Two of the most used methods for extracting structured data are wrapper induction (given a set of manually labeled
pages, a machine learning method is applied to learn extraction rules or patterns) and automatic extraction (given a
set of positive pages or given only a single page with multiple data records, generate extraction patterns).
2.3 Web Structure Mining
Links pointing to a document indicate the popularity of the document, whereas links coming out of a document
indicate the richness or the variety of topics covered in the document. Web structure mining describes the
organization of the content of the web where structure is defined by ”hyperlinks between pages and HTML
formatting commands within a page”.
Understanding the relationship between contents and the structure of the website is useful to keep an overview
about websites. One of the approach allows the comparison of web page contents with the information implicitly
defined by the structure of the website. In this way, it can be indicated whether a page fits in the content of its link
structure, and identify topics which span over several connected web pages. Thus supporting web designers by
comparing their intentions with the actual structure and content of the web page. Other studies deal with the web
page as a collection of blocks or segments. By partition the web page into blocks and by extracting the page-to-
block, block-to-page relationship from link structure and page layout analysis, a semantic graph can be constructed
over the WWW such that each node exactly represents a single semantic topic, this graph can better describe the
semantic structure of the web. Structure within a web page can be used to help machines understand pages.

961
3.Web Usage Mining Techniques
In this section, we discuss data mining techniques that are mostly used in web usage mining such as statistical
analysis techniques, clustering, classification, association rule mining, and sequential pattern mining.
Statistical analysis is the process of applying statistical techniques on web log file to describe sessions, and user
navigation such as viewing the time and length of a navigational path. Statistical prediction can also be used to
predict when some page or document would be accessed from now. It makes use of the N-grammer model which
assumes that when a user is browsing a given page, the last N pages browsed affect the probability of the next page
to be visited.
Clustering is the process of partitioning a given population of events or items into sets of similar elements. In web
usage mining there are two main interesting clusters to be discovered: usage clusters, and pages clusters. An
approach is to cluster web pages to have a high quality clusters of web pages and use that clusters to produce index
pages, where index pages are web pages that have direct links to pages that may be of interest of some group of
website navigators.
Classification is dividing an existing set of events or transactions into another predefined sets or classes based on
some characteristics. In web usage mining, classification is used to group users into predefined groups with respect
to their
navigation patterns in order to develop profiles of users belonging to a particularclass or category.
Association rule mining is the discovery of attribute values that occur frequently together in a given set of data.
Association rules mining techniques are used in web usage mining to find pages that are often viewed together, or
to show which pages tend to be visited within the same user session. A re-ranking method with the help of website
taxonomy is to mine for generalized association rules and abstract access patterns of different levels to improve the
performance of site search. Another approach for predicting web log accesses is based on association rule mining.
Association rule mining facilitates the identification of related pages or navigation patterns which can be used in
web personalization.
In sequential pattern mining a sequence of actions or events is determined with respect to time or other sequences.
In web usage mining, sequential pattern mining could be used to predict user’s future visit behaviors. Some web
usage
mining and analysis tools use sequential pattern mining to extract interesting patterns such as SpeedTracer and
Webminer.
4. Data Preprocessing for Web Usage Mining
Before data mining techniques are applied to web log file data, several preprocessing steps should be done in order
to make web log file data ready to be mined. Web log file contains data about requested URL, time and date of
request, method used, etc. The main data preprocessing tasks are data cleaning and filtering, path completion, user
identification, session identification, and session formatting.
Data cleaning is the first preprocessing task. It involves the removal or elimination of irrelevant items that are not
important for any type of web log analysis.Elimination of irrelevant items can be accomplished by checking the
suffix of the URL name to filter out requests for graphics, sound, and video hits in order to concentrate on data
representing actual page hits. For example, all log entries with filename suffixes such as gif, jpeg, and jpg can be
removed. Another cleaning process is removing log entries generated by web agents like web spiders, indexers, or
link checkers. Filtering out failed server requests, or transforming server error code is also done. Merging logs from
multiple servers and parsing the log into data fields is also considered a data cleaning step.
Path completion preprocessing task fills in page references that are missing due to local browsing caching such as
using the back button available in the browser to go back to previously visited page.
User identification is a complex step due to the existence of local caches, corporate firewalls, and proxy servers. If
the agent log shows a change in browser software, or operating system, a reasonable assumption to make is that
each different IP address in the log file represent a different user. If a page is requested that is not directly reachable
by a hyperlink from any of the pages visited by the user, a heuristic assume that there is another user with the same
IP address. Another assumption can be made is that consecutive accesses from the same host during a certain time
interval come from the same user. In some cases it is difficult to identify users, for example, when two users use the
same machine and the same browser with the same IP address and look at the same set of pages.
Session identification. A user session is defined as ”the set of pages visited by the same user within the duration of
one particular visit to a website”. Session identification is dividing the page accesses of each user into individual
sessions. One approach to identify user sessions, is by using a timeout threshold that is if the time between pages
requests exceeds a certain limit, then the user is starting a new session. Another approach assumes that consecutive
accesses within the same time period belong to the same session.
Session Formatting. A final preprocessing step could be formatting the sessions or transactions for the type of the
data mining technique, or algorithm to be applied. The Webminer, for example, formats the cleaned web server log
data in order to apply either association rule mining or sequential pattern mining.

962
5. Discussion
From previous, it is clear that making changes and adaptations to websites with the help of extracted patterns using
different data mining techniques is very effective, but doing that in the maintenance phase can be costly and time
consuming and suffers from different drawbacks. In commercial companies which are companies that sell different
kinds of products on the web, in order to make an effective maintenance to their websites, the companies have to
wait some period of time, for example one year, in order to have a representative log file that reflects customers
transactions in their website and can give a clear image about their behavior. This amount of time is considered
very big especially for the companies in which the time factor plays an important role in their success strategy, and
have many competitors who can attract their customers if they have no solid marketing strategies in order to keep
their customers as loyal as possible.
On the other hand, most businesses gather information about internet customers through online questionnaires. But,
many customers choose not to complete these questionnaires because of the amount of time required to complete
them as well as a lack of a clear motivation to complete them. Several companies use cookies to follow customers
through the WWW, but cookies are sometimes detected and disabled by web browsers and do not provide much
insight into customer preferences. This is because customers are feeling that their profiles are not secure so a
number of customers choose to give incorrect information about themselfs.
Furthermore, in web mining different strategies are implemented to identify sessions such as defining a time
threshold that a session should not exceed or assuming that consecutive accesses within the same time period
belong to the same session. In some cases, it is difficult to identify users, for example, when two users use the same
machine and the same browser with the same IP address and look at the same set of pages. We can conclude from
that, that those session and user identification strategies can not give a guarantee that those identified users and
sessions represent the actual users and sessions.
The problem of building an ill-structured website for some company/business can be solved by applying data
mining techniques such as clustering, classification, and association rule mining on the contents of the information
system of the company/business. Then, from the extracted patterns, the information needs to be considered in the
website building process is gained and invested during the design phase in the process of website design which
yields to a better designed retail website. The main advantage of this method is that it reduces maintenance time
and budgetary costs for websites if they are built taking into account the extracted interesting patterns from the
transactions database of the company/business. This approach also permits the sales manager to focus on the core
business and gives him a better view about his products and customers which is very helpful in designing retail
websites.
In conclusion, patterns extracted from applying web mining techniques on web data can be used to maintain
websites by improving their usability through simplifying user navigation and information accessibility and
improving the content and the structure of the website in a way that meets the requirements of both website owner
and user which will consequently increase the overall profit of the business.
References
1. S. Ananyan, M. Kiselev, Automated Analysis of Unstructured Texts
2. Michael Goebel, Le Gruenwald, A Survey of Data Mining and Knowledge
3. Stefan Conrad, Martin Mauve, Data Mining for Retail Website Design and Enhanced Marketing
4. Bing Liu, Web Content Mining
5. Christopher J. Hazard, Data Mining and Web Logs
6. Asem Omari and Stefan Conrad, Web Usage Mining for Adaptive and Personalized Websites
7.
Margaret H. Dunham, Data Mining Introductory and Advanved Topics
8.
Ruey-Shun Chen, Ruey-Chyi Wu, and J. Y. Chen, Data Mining Application in Customer Relationship
Management of Credit Card Business.
9.
A. McDonald and R. Welland, Web Engineering in Practice
10. Martin Ester, Hans-Peter Kriegel, and Matthias Schubert, Web Site Mining a New Way To Spot
Competitors, Customers and Suppliers in The World Wide Web.
11. M. Istrate, Web Content Mining.