A Survey on Web Page Recommendation and Data Preprocessing

© 2016, IJCERT All Rights Reserved Page | 204
International Journal of Computer Engineering In Research Trends
Volume 3, Issue 4, April-2016, pp. 204-209 ISSN (O): 2349-7084
A Survey on Web Page Recommendation
and Data Preprocessing
1
Ms. Sonule Prashika Abasaheb, 2
Prof. Tanveer I. Bagban
1 (M.E. PART-II, Department of Computer Science and Engineering, D.K.T.E. Society’s Textile
and Engineering Institute, Ichalkaranji, Shivaji University, Kolhapur, Maharashtra, India. Email:
prashikasonule@gmail.com)
2 (Associate Professor, Department of Information Technology, D.K.T.E. Society’s Textile and
Engineering Institute, Ichalkaranji, Shivaji University, Kolhapur, Maharashtra, India. Email:
tbagban@yahoo.com)
Abstract: - In today’s era, as we all know internet technologies are growing rapidly. Along with this, instantly, Web page
recommendations are also improving. The aim of a Web page recommender system is to predict the Web page or pages,
which will be visited from a given Web-page of a website. Data preprocessing is one basic and essential part of Web page
recommendation. Data preprocessing consists of cleanup and constructing data to organize for extracting pattern. In this
paper, we discuss and focus on Web page Recommendation and role of data preprocessing in Web page recommendation,
considering how data preprocessing is related to Web page recommendation.
Keywords - Recommender System, Web server logs, Web mining, Web usage mining, Data Preprocessing.
——————————  ——————————
I.INTRODUCTION
The unpredictable increase and growth of
information on the World Wide Web, with the progress
of innovative electronic devices, has made information
of Web increasingly important in everyone’s life. In
today’s era, as we all know internet technologies are
growing rapidly, so web has become large storage of
information and this amount of information grows with
high and rapid rate of change without any control of
editor; consequently, websites are also introduced
rapidly with new innovations.
Web page recommender systems can recommend Web
pages automatically which are most interesting to a
particular user based on that current Web navigation
behaviour of user. Regularly, Web users have to struggle
for finding useful pages and are very probable to leave
the site, if the index pages of a website are not well
designed. Thus, even though recommenders have been
appreciated and valued in experimental research, they
have not been commercially successful [1]. These are one
of the problems faced by user while accessing most
interested Web-pages on website.
The main reason behind these above given problems, is
the huge amount of explosive growth of information,
which is irrelevant and noisy. Considering the above
mentioned problems, it seems that there was a need of
cleaning and construct or structure that irrelevant data.
This constructing data from noisy and complex data, is
nothing but data preprocessing, we will elaborate this in
section 4.
The main problematic thing and difficulty is in accessing
most interested Web pages. Problems like this relates
with usage of Web. Hence, there is a need of cleaning
and constructing or structuring Web log data, which is
Available online at: www.ijcert.org

Sonule Prashika Abasaheb et al., International Journal of Computer Engineering In Research Trends
Volume 3, Issue 4, April-2016, pp. 204-209
nothing but data preprocessing part in Web Usage
Mining [3]. Data preprocessing plays a vigorous role
because of redundant irrelevant log data nature [4].
Thus, we find that, data preprocessing is one basic and
essential part of Web-page recommendation. This paper
is structured as below: Section 2 comprises a review of
Web page recommendation. Section 3 clarifies
categorization of recommendation system and web
mining, and it discusses how data preprocessing is
related to Web page recommendation. Section 4
illustrates data preprocessing and its steps. Section 5
provides comparative analysis of data preprocessing
techniques use; and finally, section 6 gives the
conclusion.
II. WEB PAGE RECOMMENDATION
In today’s era, along with the rapid growth of internet
technologies, there is fast increase in innovated websites.
This overpowers Web users, by offering many choices.
Therefore, Web users sometimes, probably make poor
decisions when they surf the Web. So there is a need of a
system that recommends the Web pages to Web user, to
make their work easy, such systems are nothing but Web
page recommender system.
In 1996, the term recommender system was first invented
at a workshop, and has been used inconsistently and
imprecisely in published work [1]. Web-page
recommender systems have become valuable
increasingly for helping Web users to find the most
interesting and important Web-pages on specific
websites. Best Web page recommendations can improve
website usage along with Web user satisfaction. The
important characteristic of the recommendation system is
to study from current user’s historic data and also from
remaining users. The recommendation system decides
current user’s domain from the historic data of user, then
pages prediction is done according to the domain of user
([4], [8]).
III. TAXONOMY
The recommendation system is one of the applications
exploited by results from web usage mining [4]. Thus,
Web Page recommendations are mostly related to web
mining. Therefore, we will see categorization of both Web
Page recommendation Technique and Web mining in this
section and how data preprocessing is related to Web
page recommendations.
A. Classification of recommendation technique:
The recommenders are normally implemented by
filtering algorithms categorized into three main types,
depending on how the recommendations are performed.
Thus, recommendation process or methods can be
classified as below, according to source of knowledge
used by them for recommendations ([5], [6]).
Content based recommendations: This type of systems
use, Content-Based algorithms (CB), which filter, clean
and recommend articles and items that are similar to
others accessed by the user in the past.
Collaborative Recommendations: Such process uses,
Collaborative Filtering (CF) algorithms, that clean, filter
and recommend articles and items based on the other
user preferences. For example, it recommends items
which user has not accessed in the past but, mentors of
that user liked and accessed more in the past.
Knowledge-based Recommendation: To generate a
recommendation, this method use domain knowledge
about users and items, thus need to clean users log data.
Hybrid Recommendations: This method combines
methods or techniques of two or more recommendations
from above explained categories, in order to gain better
optimization of system.
B. Classification of Web mining
A common classification or taxonomy of Web mining is
done into three different types; Web content mining, Web
structure mining and Web usage mining ([2], [3], [4]).
1) Web Content Mining: This is the process of extraction
and integration of useful information, data and
knowledge from Web page contents ([3], [10]). It is the
process to discover useful information and web page
contents like video, audio, hyperlinks as well as metadata
[2].
2) Web Structure Mining: A graph theory is used to
analyse the node and connection of a website structure in
this process. A web graph structure, contains web pages
as nodes and hyperlinks as like edges connecting pages
which are related to each other ([2], [3], [4]).
3) Web Usage Mining: This is the data mining technique
application to discover usage patterns from web data. In
our viewpoint, access logs on server side are the usage
data which keeps user navigation information [3]. We
will focus on Web Usage Mining in this section, as the
recommendation system is one of applications exploited
by results from web usage mining [4].

3.1) Web Usage Mining: Web Usage Mining is a portion
of web mining which compacts and deals with the
interesting knowledge extraction and abstraction from
log files. Usage mining tools discover and predict user
behavior, in order to help the designer to improve the
web site, such that regular users can get a personalized
and adaptive service or to attract visitors. Web Usage
Mining is also termed as Web Usage Analysis or Web Log
Mining or Analysis of Click Stream [3].
Web Usage Mining consists of three main phases. These
three phases are Data preprocessing, Pattern discovery
and Pattern analysis ([2], [3]). Among these three phases,
data preprocessing plays a vigorous role because of
redundant and noisy nature of log data. Therefore, in
next section we will focus on data preprocessing as it is
basic and essential part of Web-page recommendation.
This unit gives an overview of these three phases as
below.
Data preprocessing: It consists of cleanup and
constructing data to organize for extracting pattern. The
Data preprocessing phase includes some steps or
techniques ([2], [3]). Several data preprocessing
techniques have been used in improving other phases of
web usage mining like Pattern discovery and Pattern
analysis.
Pattern discovery: This phase deals with mining and
extraction of information from preprocessed data i.e.
results of data preprocessing phase. Techniques from
different fields such as data mining, pattern recognition
and machine learning are applied to web usage data in
order to discover user’s web access patterns [3].
Pattern analysis: This is the final stage of Web Usage
Mining. Its goal is to eliminate the irrelative patterns in
order to extract the user interesting patterns from the
results of the pattern discovery process explained above
[3]. Figure 1, below shows these Web usage mining
phases, by a basic structure [3].
Figure 1. Basic structure Web Usage Mining.
IV. DATA PREPROCESSING
According to recommendation technique categories, each
of it require to clean and filter data. Data preprocessing
converts data into a format which will be more efficiently
and easily managed user’s purpose. The main data
preprocessing task is to choose standardized data from
the initial log files, organized for algorithm of user
navigation pattern discovery [7]. This section discusses
the significance of data preprocessing methods and steps
involved in getting required content, effectively.
A] Data preprocessing steps:
Data preprocessing steps are methods or techniques that
can be applied according to source file available. These
preprocessing techniques are different for different
sources of log files. Various authors have shown that
which log file source needs which preprocessing
technique [2]. We will see this in section V below.
Various steps involved in data preprocessing phase are
shown through the figure 2; below, as given in [3]:
Figure 2. Data preprocessing steps.
1) Data Cleaning: It is a process of removing items such
as gif files, jpeg or sound files and references due to
spider navigations which are irrelevant. Data quality
which will be improved also improves analysis on data
quality ([3], [8]).
2) User Identification: This is an important step in web
usage mining of individual user identification that
Data preprocessing
Pattern Discovery
Pattern analysis
Data Cleaning
User
Identification
Session
Identification
Path Completion
Transaction
Data
Preprocessing
Web Log
File

accesses a web site. This is essential phase to determine
who has accessed Website and which pages are mostly
accessed ([3], [8]).
3) Session Identification: A sequence or series of web
pages that user browses in a single access is called as
session. The objective of session identification is to
discover and divide the page accesses of each user
session, into individual separate sessions ([3], [8], [9]).
4) Path Completion: Path completion is used to obtain
the complete user access path. This step is difficult but
important as this is final step in which user session file is
completed [3].
5) Transaction Identification: The aim of transaction
identification is to create important reference cluster for
each user. So, this is done by merging or dividing
approaches. Both approaches has a transaction list and
some parameters as an input; and a final transaction list
to be operated on by a function in the module which is
same like input format, is nothing but the output [3].
V. COMPARATIVE ANALYSIS
As shown in section two Web page recommendation
systems, use different algorithms according to the type of
recommender system. The algorithms like Content-Based
algorithms (CB) and Collaborative Filtering (CF)
algorithms are used by recommendation system based on
their type. Content based recommendations systems use,
Content-Based algorithms (CB). Collaborative
Recommendations systems use, Collaborative Filtering
(CF) algorithms, that clean, filter and recommend articles
and items based on the other user preferences ([5], [6]).
Data Preprocessing Techniques and algorithms can be
applied according to source file available. These
preprocessing techniques explained above are different
for different sources of log files. Various Authors has
shown that which log file source needs which
preprocessing technique and which algorithms can be
applied.
For example, (log file source) Server Log File needs
preprocessing techniques Data Cleaning, Log File
Filtering, Session Identification and User Identification;
and Source log file like- English Study Web site Log File
needs techniques such as Data Cleaning, User
Identification, Session Identification, Path Completion,
Transaction Identification; and algorithms, like
‘Reference Length’ and ‘Maximal Reference Length’. This
is illustrated by some examples in analysis table I as
mentioned below [2].
TABLE I. ANALYSIS FOR PREPROCESSING BASED ON
SOURCE OF LOG FILE
According to the table above, the details of algorithms
and data preprocessing implementation of web usage
mining are presented by Yan Li’s paper [10]. It explains
the reference length algorithm, which modifies the
reference length of pages in complete path by considering
Source of
log file
Preprocessing
technique
Algorithms
Applied
Authors
English Study
Web site Log
File
Data Cleaning
User
Identification
Session
Identification
Path Completion
Transaction
Identification
Maximal Forward
References(MFR),
Reference Length
Yan LI,
Boqin
FENG and
Qinjiao
MAO [10]
IIS Server Log
File
Data Cleaning
User
Identification
Session
Identification
Path Completion
Based on referred
web page and
fixed priori
threshold
Ling Zheng,
Hui Gui
and Feng Li
[11]
Web server
Log file
Data
Preprocessing
Based on
Collaborative
Filtering
JING
Chang-bin
and Chen Li
[12]
Chizhou
College
Website
Data Filtering
Session
Identification
Frame page and
Page Threshold
Fang
Yuankang
and Huang
Zhiqiu [13]

the average reference length of auxiliary pages which is
estimated and valued in advance through the maximal
forward references and reference length algorithms.
FP-Growth Algorithm was used by Huiping Peng [14] for
the web log records to be processed and a set of frequent
access patterns to be accessed. Then using both
combinations of site topology interestingness and browse
interestingness of association rules for web mining a new
pattern to provide valuable data for the site construction
was exposed.
To solve some problems that exist in traditional data
preprocessing technology for web log mining, an
improved data preprocessing technology is used by the
author ling Zheng [11]. The algorithms based on ‘fixed
priori threshold’ and ‘referred web page’ were used.
These were applied on IIS Server Log file source of log
file. Data Cleaning, User Identification, Session
Identification and Path Completion were the
preprocessing techniques applied on such sources of file.
A Web log data preprocessing algorithm, which is based
on collaborative filtering, was brought by JIANG Chang-
bin and Chen Li [12]. Even though statistic data are not
enough and records visiting user history are absent it can
perform user session identification fast, rapid and
flexibly.
Algorithms named as Frame page, Page Threshold were
applied on Chizhou College Website. Data preprocessing
techniques; Data Filtering and Session identification were
used as in [13].
VI. CONLUSION
This paper illustrated Web page recommendation and its
types to know what Web page recommendation is and
later it focuses on how data preprocessing is important
part of Web page recommendation. Thus, data
preprocessing plays a vigorous role in reducing and
removing redundant, noisy and irrelevant nature of log
data; and it is basic phase and essential for Web page
recommendation.
REFERENCES
1. Ben Schafer, Joseph A. Konstan, and John T. Riedl,
‚Recommender Systems for the Web‛.
2. Vijayashri Losarwar et al., ‚Data Preprocessing in Web
Usage Mining‛ International Conference on Artificial
Intelligence and Embedded Systems July 15-16, 2012
Singapore.
3. Naga Lakshmi et al., ‚An Overview of Preprocessing
on Web Log Data for Web Usage Analysis‛, International
Journal of Innovative Technology and Exploring
Engineering ISSN: 2278-3075, Volume-2, Issue-4, March
2013.
4. Mitali Srivastava, Rakhi Garg, ‚Preprocessing
Techniques in Web Usage Mining: A Survey‛.
5. J. Manuel Adán-Coello, C. M. Tobar, Y. Yuming,
‚Improving the Performance of Web Service
Recommenders Using Semantic Similarity‛. In JCS&T
Vol. 14, No. 2, October 2014.
6. Sabanaz S. Peerzade, Vanita D. Jadhav, ‚A Review on
Web Service Recommendation System Using
Collaborative Filtering‛, Volume 3, Issue 3, March 2015.
7. Chaoyang Xiang, Shenghui He and Lei Chen, ‚A
Studying System Based On Web Mining‛, IEEE
International Symposium On Intelligent Ubiquitous
Computing and Education, pp.433-435, 2009.
8. ‚A Survey on Preprocessing Methods for Web Usage
Data‛, (IJCSIS) International Journal of Computer Science
and Information Security, Vol. 7, No. 3, 2010.
9. R. Cooley, B. Mobasher, J. Srivastav (1999), ‚Data
preparation for mining world wide web browsing
pattern‛ in Journal of Knowledge and Data Engineering
Workshop, IEEE, Vol.1 .
10. Yan LI, Boqin FENG and Qinjiao MAO, ‚Research on
Path Completion Technique in Web Usage Mining‛, IEEE
International Symposium on Computer Science and
Computational Technology, pp. 554-559, 2008.
11. Ling Zheng, Hui Gui and Feng Li, ‚Optimized Data
Preprocessing Technology For Web Log Mining‛, IEEE
International Conference on Computer Design and
Applications, pp. VI-19-VI-21, 2010.
12. JING Chang-bin and Chen Li, ‚Web Log Data
Preprocessing Based on Collaborative Filtering‛, IEEE
2nd International Workshop on Education Technology
and Computer Science, pp.118-121, 2010.
13. Fang Yuankang et al., ‚A Session Identification
Algorithm Based on Frame Page and Page threshold‛,
IEEE Conference, pp.645- 647, 2010.
14. Huiping Peng, ‚Discovery of Interesting Association
Rules Based on Web Usage Mining‛, IEEE Conference,
pp.272-275, 2010.

AUTHOR PROFILE
1. Ms. Sonule Prashika Abasaheb is student of M.E.
PART-II, Department of Computer Science and
Engineering, D.K.T.E. Society’s Textile and Engineering
Institute, Ichalkaranji, Shivaji University, Kolhapur,
Maharashtra, India. Her interest area is Web Mining.
2. Prof. Tanveer I. Bagban is working as associate
professor in, department of Information Technology,
D.K.T.E. Society’s Textile and Engineering Institute,
Ichalkaranji, Shivaji University, Kolhapur, Maharashtra,
India. He has 14 years of Teaching Experience. His
interest area is Web Mining, Information Extraction. He
has published about 4 papers in international journals
and 2 papers in National Conferences.

A Survey on Web Page Recommendation and Data Preprocessing

More Related Content

What's hot

Viewers also liked

Similar to A Survey on Web Page Recommendation and Data Preprocessing

More from IJCERT

Recently uploaded

A Survey on Web Page Recommendation and Data Preprocessing