Download A Study on Different Types of Web Mining

Proc. of Int. Conf. on Emerging Trends in Engineering and Technology
A Study on Different Types of Web Mining
Renu, Deepti Gaur
Itm University, Gurgaon, India
Email: [email protected]
Email: [email protected]
Abstract— The enormous content of information on the World Wide Web makes it obvious
candidate for data mining research. Data mining techniques’s application to the World
Wide Web referred as Web mining where this term has been used in three distinct ways;
Web Content Mining, Web Usage Mining and Web Structure Mining.
In this survey paper, we discuss the three types of the Web mining and approaches of Web
Usage Mining and the basic Association Rules algorithm called Apriori Algorithm.
Index Terms— Web mining, Web content mining, Web usage mining, Web structure mining
I. INTRODUCTION
Web Mining is the use of the data mining techniques to automatically discover and extract information from
web documents/services. It is used to discover useful information from the World-Wide Web and its usage
patterns. The subtasks[4] of Web mining are:Resource Finding: - To retrieve intended documents and services on Web.
Information Selection/pre-processing: - Automatically extracting and pre-processing specific information
from newly discovered Web resources.
Generalization: - Refers to as the discovery of general patterns at individual Web sites and across multiple
sites.
Analysis: - The validation and/or interpretation of the mined patterns.
Visualization: - Presenting the result of an interactive analysis in a visual, easy to understand fashion.
The Three Web mining categories depends on which kind of data to be mined that is mining for information
or mining the Web link structure or mining the user navigation pattern, see Figure 1. Mining for information
focuses on the development of techniques for assisting a user in finding documents that meet a certain
criterion that is web content mining. Web content mining refers to the discovery of useful information from
web contents, which includes image, text, audio, video etc. The mining of the link structure aims at
developing techniques to take advantage of the collective judgment of web page quality which is available in
the form of hyperlinks that is web structure mining. Web structure mining tries to discover the model
underlying the link structures of the web. Model is based on the topology of hyperlinks with or without
description of links. Mining for user navigation patterns focuses on techniques which study the user behavior
when navigating the web that is web usages mining. Web usage mining refers discovery of user access
patterns from Web servers. Web usages data include data from the web server proxy server logs, browser
logs, access logs, registration data, user session or transactions, user profiles, user queries, bookmark data,
mouse clicks and scrolls, cookies, or any other data as result of interaction.
DOI: 03.AETS.2013.3.229
© Association of Computer Electronics and Electrical Engineers, 2013
A. Web Content Mining
It deals with discovering useful information or knowledge from Web pages contents. Web content is very
rich consisting of textual, images, audio, video etc and metadata(data about data) as well as hyperlinks. The
data may be unstructured (free text) or structured (data from a database) or semi-structured (html) or
Multimedia data (receive less attention than text or hypertext) although much of the Web is unstructured.
There are two approaches in web content mining that are explained as follows:
Figure 1. Web Mining Categories
Agent-Based Approach:
Agent-based web mining systems can be categories into three types as follows:
1. Intelligent Search Agents: There are several intelligent Web agents that have been developed and search
for relevant information using domain characteristics and user profiles to organize and interpret the
discovered information.
2. Information Filtering/Categorization: There are number of Web agents who use various information
retrieval techniques and characteristics o f open hypertext Web documents to automatically retrieve,
filter, and categorize them.
3. Personalized Web Agents: Personalized Web agents learn about user preferences and using this user
preference discover Web information sources, and those of other individuals with similar interests.
Database Approach
It has focused on techniques for organizing the semi-structured data on the web into more structured
collection of resources, and using standard database querying mechanisms and data mining techniques to
analyze it.
1. Multilevel Databases: In this approach the lowest level of the database contains semi-structured
information stored in various repositories of the Web (store and manage a large collection of data objects),
such as Hypertext documents. At the higher level Meta data or generalizations are extracted from lower
levels and organized in structured collections that is relational or object-oriented databases.
2. Web Query Systems: Many Web-based systems and languages utilize standard database query languages
such as SQL, structure information about Web documents, even natural language processing for the queries
that are used in World Wide Web searches.
B. Web Structure Mining
It is the Process of discovering structure information from the Web. It generates the structural summary about
the Web page and Web site and Discover the link structure of the hyperlinks at the inter-document level. It
discovers the nature of the hierarchy of hyperlinks in the website and its structure and Discovers similarities
between sites. The subtasks of Web structure mining are:
1. Finding Information about the Web page: It means to retrieve information about the relevance and the
quality of the Web page and find the authoritative on the topic and the content.
2. Inference on Hyperlink: The Web page contains not only information but also hyperlink, which contains
huge amount of annotation. Hyperlink identifies the endorsement of the author of the other Web page.
3. Authority and Hub: A hub is a page that has link to many authorities and an authority is a page with good
content on the query topic and pointed by many hub pages i.e. it is relevant and popular.
839
II. ALGORITHMS FOR WEB STRUCTURE MINING
A. HITS Algorithm
HITS that is Hyperlink-Induced Topic Search was proposed by Jon M. Kleinberg. The approach consists of
two phases:1. It uses the query terms to collect a starting set of pages (200 pages) from an index-based search engine that
is the root set of pages. The base set is generated by augmenting the root set by including all the pages to
which the root set pages link , and all the pages linking to a page in the root set, up to a designed size cutoff,
such as 3000-6000.
2. A weight-propagation phase is initiated. It is an iterative process that determines numerical estimates of
hub and authority weights.
Associate a non-negative authority weight, ap, and a non-negative weight, hp, with each p in the base set and
Initialize all a and h values to a uniform constant. The authority and hub weights are updated based on the
following equations:
ap = ∑(q such that q →p) hq
(1)
hp = ∑(q such that q ←p) aq
(2)
Equation (1) implies that if a page is pointing to by many good hubs, its authority weight should increase
(means sum of current hub weights of all the pages pointing to it). Equation (2) implies that if a page is
pointing to many good authorities, its hub weight should increase (means it is the sum of the current authority
weights of all the pages it points to). Then we define a matrix A whose rows and columns correspond to Web
pages with entry Aij=1 if page i links to page j, and 0 if not. Let a and h be the authority and hub weight
vectors, whose ith component corresponds to the degrees of authority and hubbiness of the ith page. Then :
h = A × a.
a = AT × h.
Where AT is the transposition of matrix A. Unfolding the above two equations n number of times, we have:
h = (AAT)h= (AAT)2h=------= (AAT)nh
a = (ATA)a= (ATA)2a=------= (ATA)na
B. Page Rank Algorithm
Page rank algorithm was presented and published by Sergey Brin and Larry Page at the Seventh International
World Wide Web Conference (WWW7) in April 1998. It is a search ranking algorithm that uses hyperlinks
on the Web. Based on this algorithm, the search engine Google was built, which has been a huge success.
The following ideas based on rank prestige are used to derive the Page Rank algorithm:
1. A hyperlink from a page pointing to another page is an implicit conveyance of authority to the target page.
So, the more in-links that a page i receives, the more the page i’s prestige.
2. Pages that point to page i also have their own prestige scores. A page with a higher prestige score that
points to i is more important than a page with a lower prestige score pointing to i. Or we can say, a page is
important if it is pointed to by other important pages.
According to rank prestige in social networks, the importance of page i (i’s Page Rank score) is determined
by summing up the Page Rank scores of all pages pointing to i. Since a page may point to many other pages,
the prestige score of this page should be shared among all the pages that it points to. Suppose Web as a
directed graph G = (V, E), where V is the set of vertices or nodes that is the set of all pages and E is the set of
directed edges in the graph that is hyperlinks. Let the total number of pages on the Web be n (that is, n =|V |).
The Page Rank score of the page i (denoted by P(i))is defined by
P(i) = ∑ P(j)/Oj,
(j,i)є E
Where Oj is number of out-links of page j.
Let P be a n-dimensional column vector of Page Rank values that is P = (P(1),P(2),...,P(n))T. Let A be the
adjacency matrix of our graph with
1/Oj if (i, j) є E
Aij = 0 otherwise
Or the system of n equations with
P = ATP
( 1)
Which is the characteristic equation of eigen system, where the solution to P is an eigenvector with the
corresponding eigen value of 1.Since this is a circular definition, we use an iterative algorithm is to solve it.
The Power Iteration Method for Page Rank
840
Page Rank-Iterate(G)
P0 ← e/n
k←1
repeat
Pk ← (1 − d )e + dAT Pk -1;
k ← k + 1;
until ||Pk – Pk-1||< ε
return Pk
The problem is that Eqn 1 does not quite suffice because the Web graph does not meet the conditions. In fact,
Eqn 1 can also be derived based on the Markov chain. After augmenting the Web graph to satisfy the
conditions, the following Page Rank equation is produced:
P = (1 − d)e + dATP
where e is a column vector of all 1’s. So the Page Rank formula for each page i:
n
P(i) = (1-d)+d∑ Aji P(j)
j=1
and it is equivalent to the formula given in the original Page Rank papers:
P(i) = (1-d)+d ∑ P(j)/Oj,
(j,i) є E
The parameter d is called the damping factor which can be set to a value between 0 and 1.
C. Web Usage Mining
Discovery of meaningful patterns[2],[5] from data, generated by the client-server transactions on one or
more Web servers. Typical Sources of the Data are:
1. Automatically generated data that was stored in server access logs(records all requests that are processed
by the server), referrer logs, agent logs, and client-side cookies.
2. E-commerce and product-oriented user events (e.g. shopping cart changes, ad or product click-throughout
etc.)
3. User profiles and/or user ratings
4. Meta-data, page attributes, structure of each page contained in a web site, content of the web page.
Web Log Format: A Web server log file contains request made to the web server. The most popular log
format are Common Log Format (CLF) and the extended CLF. CLF file is created by the web server to keep
track of the requests that occur on a web site. The CLF is shown as follows:<ip_addr><base_url>_<date><method><file><protocol><code><bytes><referrer><user_agent>
Approach of Web usage mining:
The web usage mining includes the following several steps: data collection, data pretreatment, knowledge
discovery and pattern analysis.
Data collection: The first step in Web usage mining and consist of collecting the relevant web data which
will be analyzed to provide useful information about the user’s behavior. Two main sources of data for Web
usage mining are Data on the Web server side and Data on the client side. And if intermediaries introduced in
the client server communication, then they also become source for usage data such as Proxy Servers and
packet sniffers.
Data Preprocessing: Some databases are inconsistent, insufficient and noisy. Data Preprocessing[7] or
pretreatment is used to transform those databases into integrate and consistent database which may be mine.
It includes User Identification, Data Fusion and Cleaning, Session Identification and Path completion. Web
usage mining process is incomplete without using preprocessing section.
Data Fusion and Cleaning: In some cases, multiple servers are used to reduce the load on any particular
server. So, merging of log files from several Web and application server is known as Data Fusion, see Figure
2. A user comes from multiple Web or application servers then data fusion merge data and solved various
user identification session etc.
841
Figure 2. Data Fusion
Data Cleaning[2] is used to eliminate irrelevant items means irrelevant records in web access log to be
eliminated. Data cleaning is mostly use to removing extraneous references to embedded objects that may
not be important for the purpose of analysis, including references to graphics, style files, or sound files.
Some information should not provide useful information in analysis or data mining tasks then Data cleaning
is us ed. Remove erroneous references.
The following two kinds of records are unnecessary and should be removed:
1. The records of graphics, videos and the format information.
2. The record with the failed HTTP status code like record with the status codes over 299 or under 200
should be removed.
Algorithm for Data Cleaning
Read record in database
For each record in database Read fields (URI-Stem)
If fields = {*.gif, *.jpg, *.css} then
Remove records
Else
Save records
End if
Next record
User Identification: In web usages mining does not require knowledge about a user history because the users
visit or request given more than one time to the server. If we visit more than one time, then it generate
multiple sessions for each user. It is used to identify who access the Web site and which page are accessed.
It is also known as user activity records. User is identified by using IP address and user agent in log files.
Client request to the server, then it generates log files and at that time client also send the user agent to the
server. It can be explained with the help of following example:
TABLE I: USER DATA
TIME
0.04
0.10
0.12
0.15
0.20
0.25
0.28
0.33
0.35
IP
192.168.100.101
192.168.100.101
192.168.100.102
192.168.100.102
192.168.100.102
192.168.100.102
192.168.100.101
192.168.100.101
192.168.100.102
URL
A
B
A
B
C
D
C
D
D
REFF
A
A
B
C
B
C
C
AGENT
X
X
Y
Y
Y
Y
X
X
Y
Then, the user can be find out by the combination of IP + AGENT using Table 1 as follows:
TABLE II: USER1
TIME
0.04
0.10
0.28
0.33
IP
192.168.100.101
192.168.100.101
192.168.100.101
192.168.100.101
URL
A
B
C
D
842
REFF
A
B
C
AGENT
X
X
X
X
T ABLE III: U SER 2
TIME
0.12
0.15
0.20
0.25
0.35
IP
192.168.100.102
192.168.100.102
192.168.100.102
192.168.100.102
192.168.100.102
URL
A
B
C
D
D
REFF
A
B
C
C
AGENT
Y
Y
Y
Y
Y
The Table 2 and Table 3 shows 192.168.100.101 visits more than one time as well as 192.168.100.102 also
visits. So, we find the users by the IP and User Agent.
Session Identification: It is used to divide the page access of each user at a time into individual sessions. A
session is a sequence of page views by a single user during a single visit. A session is the process of user
activity record of each user in the log file and is Used to find that how many session create a single user login
to the Web site. Session can be used in two ways: 1. Time Oriented
2. Structure Oriented
Time Oriented: Time Oriented depends on the timestamp or date and time of request in the server log file. In
the time oriented session there are two types: i) The difference between the First and Last request is <= 30 minutes.
ii) The difference between First request and next request is <=10.
And it is called as one session.
Structure Oriented: Structure Oriented capture in the referrer field of server log and depends on referrer field
is currently open or that user currently login referrer. Means it is belonging to more than one “open”
constructed session as shown in Table 4.
For example: TABLE IV: SESSION DATA
TIME
0.04
0.10
0.12
0.15
0.20
0.25
0.48
0.52
0.58
IP
192.168.100.101
192.168.100.101
192.168.100.102
192.168.100.102
192.168.100.102
192.168.100.102
192.168.100.101
192.168.100.101
192.168.100.102
URL
A
B
A
B
C
D
C
D
D
REFF
A
A
B
C
B
C
C
AGENT
X
X
Y
Y
Y
Y
X
X
Y
In Table 5 structure oriented session means 192.168.100.102 this user session open for time s tamp 0:12 to
0:25 and 0:58. It is consider as one session.
T ABLE V: SESSIONS WITH STRUCTURED ORIENTED
TIME
0.12
0.15
0.20
0.25
0.58
IP
192.168.100.102
192.168.100.102
192.168.100.102
192.168.100.102
192.168.100.102
URL
A
B
C
D
D
REFF
A
B
C
C
AGENT
Y
Y
Y
Y
Y
TABLE VI: SESSION1
TIME
0.12
0.15
0.20
0.25
IP
192.168.100.102
192.168.100.102
192.168.100.102
192.168.100.102
URL
A
B
C
D
REFF
A
B
C
AGENT
Y
Y
Y
Y
TABLE VII: SESSION 2
TIME
0.58
IP
192.168.100.102
URL
D
843
REFF
C
AGENT
Y
Path completion: After completion sessions as shown in Table 6 and Table 7 we start path completion,
because that user how web pages visited that should be confirmed using path completion phase. Some
reasons that causes Path incompletion are:- Local cache, agent cache and browser’s back button that can
result in some important access not recorded in the access log file and the number of URL recorded in log
may be less than the real one. To discover user’s travel pattern, the missing page in the user access path
should be appended. In path completion missing Reference method is used means users backtrack should
not be stored in server log file.
Pattern Discovery: Use statistical method to carry on the analysis such as frequency analysis, mean,
median, etc. and mine the pretreated data[1]. Clustering of users help to discover groups of users with
similar navigation patterns (provide personalized Web content). Classification is the technique to map a
data item into one of several predefined classes. Association Rules discover correlations among pages
accessed together by a client. Sequential Patterns extract frequently occurring inter- session patterns such
that the presence of a set of items s followed by another item in time order. Dependency Modeling
determines if there are any significant dependencies among the variables in the Web.
Association Rule: Let I = {i1, i2…………. In} be a set of n different items, and DB = {T1, T1…………… Tm} be the
transaction database consisting of m transaction where each transaction Ti = {i1, i2…………..ik} is a set of k
elements from I.
An association rule [ 3] is then specified as X=>Y, where X, Y belongs to I and X∩Y= Ø. Such rules have
two attributes associated with them that are Support and Confidence.
Let S be the % of transaction in DB which contains XᴜY then S is known as Support of X=>Y. Let C be the
% of transaction in DB containing X which also contain Y then the rule X=>Y holds with confidence C. X is
known as antecedent and Y as consequent.
Association Rule in Web usage mining: In Web usage mining, association rule[6] refers to set of pages which
are accessed together with a minimum support value which can help in organizing Web space efficiently. For
example:- Consider if 70% of the users who accessed get/programs/courses/x.asp also accessed
get/programs/courses/y.asp, but only 30% of those who accessed get/programs/courses accessed
get/programs/courses/y.asp, then it shows that some information in x.asp is making the clients access y.asp.
This helps the designers to decide on designing a link between the above two pages. So, Association rule
mining is one of the most popular pattern –discovery methods.
Steps of Apriori Algorithm: 1. By scanning the database iteratively to find the support count of each k-itemset
where k=0,1,….p such that p≤ maximum number of items in the database. All those item sets whose support
count is greater than or equal to the user specified minimum support is known as a frequent itemset.
2. Generate association rules from the frequent item sets. For every frequent itemset X, if X,Y≠ Ø and
support(X)/support(Y) ≥minimum confidence, then Y=>(X-Y)
Apriori Algorithm: Find the frequent itemsets that uses an iterative level-wise approach based on candidate
generation,.
Input : Database D of transactions; minimum support threshold, min_sup.
Output : L, frequent itemsets in D.
Method :
C1 = {Candidate 1-itemsets};
L1 = {c є C1 | c.count ≥ minsupport};
For(k=2;Lk-1 ≠ Ø;k++) DO Begin
Ck = Candidate-gen(Lk-1);
For all transactions Ti є DB Do Begin
Ct = subset(Ck,t);
For all candidate c є Ct Do
c.count++;
End
Lk = {c єCk | c.count ≥ minsup}
End
Answer = ᴜkLk;
Procedure Candidate-gen(Lk-1)
Ck ← Ø;
For all l1,l2 є Lk-1
With l1 = {i1,……..ik-1}
844
l2 = {i1,…….i’k-1}
And ik-1 < i’k-1 Do
C ← {i1,….ik-1, i’k-1};
Ck ← Ck ᴜ {c}
For each (k-1) subset s of c do
If(s є Lk-1) then
Delete c from Ck;
End
End
Return Ck;
FP Growth Algorithm: FP-growth algorithm is the first pattern-growth concept algorithm. FP-growth
algorithm constructs an FP-tree structure and mines frequent patterns by traversing the constructed FP-tree.
An FP-tree is a compact data structure that represents the data set in tree form.
FP-tree structure :The FP-tree structure[8] has sufficient information to mine complete frequent patterns. It
consists of a prefix tree of frequent 1-itemset and a frequent-item header table. A node in the prefix-tree has
three fields: item-name (register which item this node represents), node-link and count.
1. Item-name is the name of the item.
2. Node-link is the link to the next same item name node in the FP-tree. Each entry in the frequent-item
header table has two fields: item-name and head of node-link.
3. Head of node-link is the link to the first same item-name node in the prefix-tree.
4. Count is the number of transactions that consist of the frequent 1-items on the path from root to this
node.
Construction of FP-tree
1. First, create the root of the tree, labeled with “null”.
2. Scan the database D again. (Before this, we scanned it to create 1-itemset and then L).
3. The items in each transaction are processed in L order (i.e. sorted order).
4. A branch is created for each transaction with items having their support count separated by colon.
5. Whenever the same node is encountered in another transaction, we increment the support count of the
common node or Prefix.
6. To facilitate tree traversal, an item header table is built such that each item points to its occurrences in
the tree via a chain of node-links.
7. Now, The problem of mining frequent patterns in database is transformed to that of mining the FPTree.
Mining the FP-Tree by Creating Conditional (sub) pattern bases
1. Start from each frequent length-1 pattern (as an initial suffix pattern).
2. Construct its conditional pattern base which consists of the set of prefix paths in the FP-Tree cooccurring with suffix pattern.
3. Then, Construct its conditional FP-Tree & perform mining on such a tree.
4. The pattern growth is achieved by concatenation of the suffix pattern with the frequent patterns
generated from a conditional FP-Tree.
5. The union of all frequent patterns (generated by step 4) gives the required frequent itemset.
Advantages of FP-Growth Algorithm
1. FP-growth is an order of magnitude faster than Apriori algorithm and tree-projection.
2. No candidate generation, no candidate test
3. Use compact data structure
4. Eliminate repeated database scan
Pattern Analysis: Pattern Analysis is a final stage of the whole Web usage mining. The main goal of this
process is to eliminate irrelevant (or unwanted rules) or patterns and to extract the interesting rules or
patterns from the output of the pattern discovery process. The output of earlier stage web usage mining is often
not suitable for the web site administrators. The type of information sought according to this is “How are people
using the sites? Which pages are being accessed most frequently?” All these type of questions require the
analysis of the structure of hyperlinks as well as the contents of the pages. This can be done with the help of
some analysis methodologies and tools. The common techniques which have been used for pattern analysis
are visualization techniques, Data & Knowledge Querying, OLAP(online analytical processing) techniques,
and Usability Analysis.
845
Visualization Techniques: Visualization has been used very successfully in helping people understand
various types of the phenomena, both real and abstract. Hence it is a natural choice for understanding the
behavior of web users. Groth (1999) argues that visualization is simply the graphical presentation of data.
OLAP Techniques: On-line Analytical Processing (OLAP) is emerging as a very powerful paradigm for
strategic analysis of databases in business settings. The key characteristics of strategic analysis include:
very large data volume, support for various kinds of information aggregation, explicit support for the
temporal dimension, and long-range analysis in which overall trends are more important than details
of individual data items.
On-line Analytical Processing can be performed directly on top of relational databases, but industry
has developed specialized tools to make it more efficient and effective.
Data and Knowledge Querying
The reason for the great success of relational database technology has been the existence of a high-level,
declarative, query language, which allows an application to express what conditions must be satisfied by
the data it needs, rather than having to specify how to get the required data.
Given the large number of patterns that may be mined, there appears to be a definite need for a
mechanism to specify the focus of the analysis. Firstly, constraints may be placed on the database to
restrict the portion of the database from which to mine for. And secondly, querying may be performed on the
knowledge that has been extracted by the mining process, and in this case a language for querying
knowledge rather than data is needed.
Usability Analysis: The first step undertaken in this method is to develop instrumentation methods that
collect data about Software usability. This data is then used to build computerized models and simulations
that explain the data. Finally, various visualization and data presentation
techniques are used to
help an analyst understand the phenomenon. And usability analysis approach can also be used to model
the browsing behavior of users on the web.
However, as most of those techniques are disliked by users because of slow speeds, inflexibility, difficult to
maintain and limited functionality. There still remains a lot of work to be undertaken by both researcher and
developer to develop a more efficient, flexible and powerful set of tools to undertake this task.
III. CONCLUSION
Web Mining is used to extract the useful information from very large amount of web data. In this
article, we tried to give a brief introduction to the broad field of Web mining. Therefore, we motivated this
field of research, gave a more formal definition of the terms used herein and presented a brief overview of
currently available Web mining techniques and preprocessing and discovery of Web usage mining.
REFERENCES
[1] Dr. Punithavalli and Mrs. V. Sujatha, “A Study of Web Navigation Pattern Using Clustering Algorithm in Web Log
Files,” IJSER, Vol.- 2, Issue – 9, September – 2011.
[2] Mr. Balram Purswani and Mr. Akshay Upadhyay, “ Web Usage Mining has Pattern Discovery,” IJSER, Vol.- 3,
Issue – 2, ISSN 2250-3153, February – 2013.
[3] Ms Kiruthika M, Ms Dipa Dixit and Mr. Rahul Jadhav, “ Pattern Discovery Using Association Rules,” IJASCA,
Vol. – 2, No.- 12, 2011.
[4] Hemant Kumar Singh and Brijendra Singh, “Web Data Mining Research: A Survey,” IEEE – 2010.
[5] Jaideep Srivastava, Robert Cooley, Mukund Deshpande and Pang-Ning Tan, “Web Usage Mining: Discovery and
Applications of Usage Patterns from Web Data,’ SIGKDD Explorations, Vol.- 1, Issue – 2, Jan.-2000.
[6] Prof. Marathe Dagadu Mitharam, “Association Rule in Web Usage mining,” IJSER, Vol.- 3, Issue- 4, April – 2012.
[7] Prof. Marathe Dagadu Mitharam, “ Preprocessing in Web Usage Mining,” IJSER, Vol.- 3, Issue – 2, ISSN 22295518, February – 2012.
[8] R. Ivancsy, I. Vajk, “Frequent Pattern Mining in Web Log Data” Acta Polytechnica Hungarica Vol. 3, No. 1, pp. 7790, 2006.
846