AN INTELLIGENT OPTIMAL GENETIC MODEL TO INVESTIGATE THE USER USAGE BEHAVIOUR ON WORLD WIDE WEB

International Journal of Data Mining & Knowledge Management Process (IJDKP) Vol.3, No.2, March 2013

AN INTELLIGENT OPTIMAL GENETIC MODEL TO
INVESTIGATE THE USER USAGE BEHAVIOUR ON
WORLD WIDE WEB
V.V.R. Maheswara Rao1 and Dr. V. Valli Kumari2
1
Professor, Department of Computer Applications,
Shri Vishnu Engineering College for Women, Bhimavaram, Andhra Pradesh, India,
mahesh_vvr@yahoo.com
2
Professor, Department of Computer Science & Systems Engineering,
College of Engineering, Andhra University, Visakhapatnam, Andhra Pradesh, India,
vallikumari@gmail.com

ABSTRACT
The unexpected wide spread use of WWW and dynamically increasing nature of the web creates new
challenges in the web mining since the data in the web inherently unlabelled, incomplete, non linear, and
heterogeneous. The investigation of user usage behaviour on WWW is real time problem which involves
multiple conflicting measures of performance. These measures make not only computational intensive but
also needs to the possibility of be unable to find the exact solution. Unfortunately, the conventional methods
are limited to optimization problems due to the absence of semantic certainty and presence of human
intervention. In handling such data and overcome the limitations of conventional methodologies it is
necessary to use a soft computing model that can work intelligently to attain optimal solution.

To achieve the optimized solution for investigating the web user usage behaviour, the authors in the present
paper proposes an Intelligent Optimal Genetic Model, IOGM, which is designed as an optimization tool
based on the concept of natural genetic systems. Initially, IOGM comprise a set of individual solutions or
chromosomes called the initial population. Later, biologically inspired operators create a new and
potentially better population. Finally, by the theory of evolution, survive only optimal individuals from the
population and then generate the next biological population. This process is terminated as when an
acceptable optimal set of visited patterns is found or after fixed time limit. Additionally, IOGM strengthen
by its ability to estimate the optimal stopping time of process. The proposed soft computing model ensures
the identifiable features like learning, adaptability, self-maintenance and self-improvement. To validate the
proposed system, several experiments were conducted and results proven this are claimed in this paper.

KEYWORDS
Web usage mining, Genetic Algorithm, Optimal Solution, Selection, Crossover, Mutation

1. INTRODUCTION
The rapid advances in data generation, availability of automated tools in data collection and
continued decline in data storage cost enable with high volumes of data. In addition, the data is
non scalable, highly dimensional, heterogeneous and complex in its nature. This situation creates
inevitably increasing challenges in extracting desired information. Thus, web mining evolves into
a fertile area and got the focus by many researchers and business analysts. Web mining is a
methodology that blends conventional techniques with incremental algorithms. Web mining is the

DOI : 10.5121/ijdkp.2013.3203 33

application of data mining techniques to discover and retrieve useful information and patterns
from the WWW documents and services.

The web mining can be used to discover hidden patterns and relationships within the web data.
The web mining task can be divided into three general categories, known as Web Content Mining
(WCM), Web Structure Mining (WSM) and Web Usage Mining (WUM) as shown in figure 1.

Figure 1. Types of web mining

Web content mining is the mining technique that can extract the knowledge from the content
published on internet, usually as semi-structured (HTML), unstructured (Plain text) and structured
(XML) documents. The content of a web page may be varied, like text, images, HTML, tables or
forms. The ability to conduct web content mining allows results of search engines to maximize
the flow of customer clicks to a web site, or particular web pages of the site, to be accessed
numerous times in relevance to search queries. The main uses of this web mining are to gather,
categorize, organize and provide the best possible information.

Web structure mining extracts the knowledge from the WWW and links between references in the
web. Mining the structure of the web involves extracting knowledge from the interconnections of
the hypertext documents in the WWW. This results in discovery of web communities, and also
pages that are authoritative. The main purpose for structure mining is to extract previously
unknown relationships between web pages.

Web usage mining is the process of automatic discovery and investigation of patterns in click
streams and associated data collected or generated as a result of user interactions with web
resources on web sites. The main goal of web usage mining is to capture, model and analyze the
behavioural patterns and profiles of users interacting with web sites. The discovered patterns are
usually represented as collection of pages, objects or resources that are frequently accessed by
groups of users with common needs or interests. The primary data resources used in web usage
mining are log files generated by web and application servers. The overall Web usage mining
process can be divided into mainly three interdependent stages: Pre-processing, Pattern Discovery
& Pattern Analysis as shown in figure 2.

Figure 2. Stages of web usage mining

34

Pre-processing is Initial and very important stage in web usage mining applications in the creation
of suitable target data set to which mining algorithms can be applied. The inputs for pre
processing stage are the Web Server Logs, Referral Logs, Registration Files and Index Server
Logs. In the pre processing stage, the click stream data is cleaned and portioned into a set of user
transactions representing the activities of each user during different visits to the sites. The pre
processing stage ultimately results in a set of user sessions, each corresponding to a delimited
sequence of page views. However, in order to provide the most suitable data for further stages of
web usage mining, Pre processing is an aspect of data mining whose importance should not be
underestimated. Unless this phase is performed adequately, it is not possible for mining
algorithms to provide reliable results.

Pattern discovery is the second stage of web usage mining process which can take the output
generated by pre processing stage. The goal of pattern discovery is the stage of learning some
general concepts from the pre processed data. In this phase, statistical, data base and machine
learning techniques like classification, clustering and association rule mining are performed on
the extracted information. This stage obtains possibly hidden patterns reflecting the typical
behaviour of users, as well as summary statistics on web resources, sessions and users.

Pattern analysis is the final stage of usage mining which can extract interested patterns from the
output of pattern discovery. The goal of pattern analysis is the task of understanding, visualizing,
and interpreting the discovered patterns and statistics. The most common approaches for the
pattern analysis are knowledge query mechanism such as SQL and OLAP. Any method used in
the pattern analysis stage assumes the output of the previous phase that has already been
structured. The output generated by pattern analysis is used as input to the various applications
such as recommendation engines, visualization tools and web analytics. The types and levels of
analysis, performed on the integrated usage data, depend on the ultimate goals of the analysts and
the desired outcomes.

Due to the dynamic and unrealistic environment of web log data, the existing systems find
difficulty in handling the newly emerged problems throughout all the phases of web usage mining
especially, the pattern discovery. To proceed towards web intelligence, reducing the need of
human intervention, it is necessary to integrate and entrench artificial intelligence into web
mining tools. To achieve the intelligence, soft computing methodologies seem to be a good
candidate.

The soft computing models are characterized by its ability for granular computation in avoiding
the concept of approximation. Basically, soft computing models provide the foundation for
computational intelligence systems and further outline the basis of future generation computing
systems. These models are close resemblance to human like decision making and used for
modelling highly non linear data, where the pattern discovery, rule generation and learnability are
typical. The Fuzzy Logic, Artificial Neural Networks, Genetic Algorithms and various
combinations of these techniques have made the Soft computing paradigm. Among which,
Genetic Algorithms, a biologically inspired technology and is more suitable for web data, which
is intrinsically unlabelled, heterogeneous and dynamic.

Genetic algorithms are examples of evolutionary computing methods and are of optimization type
algorithms for both supervised and unsupervised techniques. The Genetic Algorithm is an
elegant, simple, yet extremely powerful and adoptive model in resulting exact optimal solutions.
The GA is more adequate since the implicit parallelism of GA can mine the large web data in less
time and the stochastic ability of GA yields in optimum solution. Moreover, the GAs minimize
the assumptions in making the real study of web user usage behaviour. Thus, in order to find the
optimized solution for investigating the web user usage behaviour, it is essential to make use of
Genetic Algorithms with granular computing nature.
35

The goal of present paper is to deploy Intelligent Optimal Genetic Model IOGM, in order to find
the optimized solutions for investigating the web user usage behaviour. IOGM is modelled on the
principles of natural genetic systems, where the genetic information of each individual is encoded
in structures called chromosomes. Each chromosome has an associated fitness value, which
indicates its degree of goodness with respect to the solution it represents. Biologically inspired
operators like selection, crossover and mutation are applied on the chromosomes in the
population to yield potentially optimal solutions.

This paper is organized as follows. In section 2, related work is described. In next section 3,
proposed work is presented in detail. In subsequent section 4, the experimental analysis of
proposed work is shown. Finally in section 5 conclusions are mentioned.

2. RELATED WORK
The literature survey is conducted by the authors for the proposed work from 1999 to current year
that comprises of major advances in the field of molecular biology, coupled with advances in
genomic technologies have lead to an explosive growth in biological information generated by
scientific community. It is an interdisciplinary field involving Biology, Computer Science,
Mathematics and Statistics to analyze biological sequence data.

In the year 1999, M. Martin-Bautista and M. A. Vila [13] reviewed the genetic features in the data
mining issues. They expressed that application of soft mining techniques to data mining and
knowledge discovery is necessary in order to enhance the effectiveness of traditional data mining
techniques. And they presented the features of genetic algorithms and its parameters.

In the next year 2000, H. Kargupta and S. Bandyopadhyay [7] explored the importance of linkage
learning in genetic algorithms and other optimization algorithms. Linkage learning deals with the
issue of intelligence acquired by the properties of representation. They developed the foundation
to implement the genetic algorithms. They also identified the future directions of linkage learning
in genetic algorithms.

S. Bandyopadhyay and S. K. Pal [22] presented in 2001 the analogy of concept of chromosome
differentiation and variable string length in genetic algorithms. They also explained the
integration of genetic algorithms to the classification problems. They demonstrated the
effectiveness of GA with a variable string length classifier – VGACD.

During 2002, F. Picarougne, N. Monmarche, A. Oliver, and G. Venturini [4] presented a genetic
search strategy for a search engine problem. They have defined search strategy which implements
as directly as possible the concepts used in evolutionary algorithms. Their experiments have
shown the relevance of genetic approach in search engine problems. They used a standard GA
and they have directed towards advance algorithms with additional evolutionary techniques in
their perspectives. Freitas, A.A [5] they discussed the use of evolutionary algorithms, particularly
genetic algorithms in data mining and knowledge discovery. They also focused in the data mining
task of classification. They felt that design evolutionary algorithms are a promising research
direction in the knowledge discovery process as mentioned in their future research directions. H.
K. Tsai, J. M. Yang, and C. Y. Kao. [8] demonstrated the usage of GA in finding the optimal
global strategies by using clustering technique on biological datasets. Sankar K. Pal, Fellow [21]
a deep survey of the existing literature on soft web mining along with commercially available
systems. They also summarized the different types of web mining and its components. They
illustrated the relevance of soft computing including GAs with examples. The extensive
bibliography provided by them is evidence the relevance of usage of GAs in web mining. Sankar
K. Pal, Fellow [30] clearly mentioned that the incorporation of computational intelligence
techniques for mining the web is a future research direction.
36

Within the year 2003, B. Minaei-Bidgoli, William F. Punch [1] presented an approach for
classifying the students in order to predict their final grade based on features extracted from log
data of an education web based system. To minimize the prediction error rate they used genetic
algorithms by weighting the features. They profoundly expressed applying the evolutionary
algorithms for association rules to improve prediction accuracy as their future work. C. Lopez-
Pujalte, V. P. G. Bote, and F. de Moya Anegon [3] evaluated the efficacy of genetic algorithm for
various order based fitness functions that are designed as measures of goodness of the solutions.
K. C. Wiese and E. Glen [12] presented a genetic algorithm to predict the secondary structure of
RNA molecule where secondary structure is encoded as a permutation. More specifically, they
focused at selection strategies, crossover operators and selection procedures. They also provided
comparison study of several crossover operators.

All the range in 2004, B. Minaei-Bidgoli, G. Kortemeyer, and W. F. Punch [2] extended their
previous work [1] and presented a new approach for predicting students’ performance based on
extracting the average of feature values for overall of the problem. Romao, W., Freitas, A. and
Gimenes I [20] proposed a genetic algorithm designed specifically to discover of knowledge in
form of prediction. A prototype was implemented and applied to a real world science and
technology database. In addition, it works with natural linguistic terms which lead to the
discovery of more comprehensible knowledge. S. Bandyopadhyay, S. K. Pal, and B. Aruna [23]
developed a pattern classification methodology based on multi objective variable length GA to
solve the problem of generating class boundaries. The usage of this classifier in other domains is
a future work. S. J. Louis and J. McDonnell [24] provided a frame work for evaluating machine
learning systems. S. Hill, J. Newell, and C. O'Riordan [25] proposed an inversion operator in a
basic genetic algorithm and also compare the effectiveness. The results shown by them indicated
that inversion operator is more useful in genetic algorithms with fitness scaling.

S. Y. Wang and K. Tai [28] implemented a bit array representation method for structural topology
optimization using the genetic algorithm in 2005. An identical initialization method is also
proposed to improve the GA performance.

In the subsequent year 2006, S. Y. Wang, K. Tai, and M. Y. Wang [29] presented a versatile,
robust and enhanced genetic algorithm for structural topology optimization using problem
specific knowledge. In their implementation process specifically pronounced the importance of
choosing appropriate representation techniques, genetic operators and evaluation methods.

During 2008 and 2009, S. Ventura, C. Romero, A. Zafra, J. A. Delgado, and C. Hervas [27]
designed a framework that can apply to maximize reusability and availability of evolutionary
computation with a minimum effort in web mining. They also implemented a graphical user
interface Gen Lab that allows users to configure an algorithm, to execute it interactively and to
visualize the results obtained. The heavily demanding computational performance is an open
problem as enmarked in their future research work.

The present IOGM frame work concentrating on assigning the weights to the characteristics of
initial population, which minimizes the error rate in identifying the relation among the pages.
Data analysis tools used earlier in web mining were mainly based on statistical techniques like
regression and estimation. Recently GAs have been gaining the attention of researchers for
solving certain web mining problems with the need to handle large data sets in computationally
efficient manner.

37


3. PROPOSED INTELLIGENT OPTIMAL GENETIC MODEL: IOGM
While investigating the web user usage behaviour, the web miners have to take many intelligent
,
decisions at each stage of web user usage mining. All such decisions have to be made sequentially
.
at different levels within time. Since the web log data impacted by many external and explicit
functions, therefore the data in the web log becomes non linear and the problem space turn into so
dynamic. Consequently, the programming techniques used for this type of the problem happen to
techniques
computationally expensive. To attain global solution and minimize the efforts required, the
present chapter proposes an Intelligent Optimal Genetic Model IOGM which is a standard non
linear optimization technique.

Optimization is the method of obtaining the global solution under any given circumstances. There
are many optimization techniques in the literature

• Classical
• Linear programming
• Non Linear Programming with & without constraints
• Dynamic Programming
• Stochastic Programming
• Soft computing Techniques

Out of which a the soft computing technique, Genetic Algorithm used for IOGM, which is well
echnique,
suited to solve the problem of investigation of web user usage behaviour since the weblog data
characterized by both continuous and discontinuous functions. To get the optimum solution for
th
investigating the web user usage behaviour it is necessary to be processed the potential patterns
extracted by pattern discovery technique of web usage mining process.

The proposed work considers each step of Genetic Algorithm in the light of web mining. Towards
eb
this goal, the authors present the IOGM, which equally concentrates on all steps of genetic
eps
process and more adaptable to incremental web log scenario. The architecture and the process of
IOGM are as shown in Figure 3.

Figure 3. Architecture of IOGM

The first and perhaps the most typical task is to identify the initial population by determining as a
set of individual solutions from the list of visited patterns to initiate the process using IOGM
ate
38

encoding strategy. Subsequently, by the theory of evolution, only optimal individuals survive
from the population and then generate the next biological population based on IOGM fitness
function. In the next step, biologically inspired IOGM operators create a new and potentially
better population. Later, IOGM sorts the visited patterns in desired order based on past
performance combined from human thresholds. Finally, this process is terminated as and when
an acceptable optimal set of visited patterns is found or after fixed time limit of IOGM end
function.

3.1 IOGM encoding strategy – Initial population:
The encoding strategy is a process of representing the potential solution to a problem into a
suitable form of viable individuals so that the genetic algorithm can process. It is crucial issue in
genetic process as it plays a critical role to arrive at best performance of algorithm as robust as
possible. Various encoding strategies have been introduced in the literature for effective
implementation of genetic algorithms. IOGM adopts the Binary encoding strategy to determine
initial population from the visited patterns generated by pattern discovery stage of web usage
mining process.

The sample output generated by pattern discovery of web usage mining process is as shown in
figure 4. This can be considered as input for IOGM.

Ending with P5: {P5}, {P4, P5}, {P1, P4, P5}, {P3, P5}, {P1, P5}
Ending with P4: {P4}, {P3, P4}, {P2, P3, P4}, {P1, P3, P4}, {P2, P4}, {P1, P2, P4}, {P1, P4}
Ending with P3: {P3}, {P2, P3}, {P1, P2, P3}, {P1, P3}
Ending with P2: {P2}, {P1, P2}
Ending with P1: {P1}
Figure 4. Sample visited web pages
The encoding strategy adopted by IOGM creates various sets of chromosomes with genes as
possible solution. The chromosome set is strings of 0’s & 1’s and coded as finite-length string
over an alphabet of finite length. A commonly used principle for coding is known as principle of
minimum alphabet [16]. Each chromosome refers to a coded possible solution. A set of such
chromosomes in a generation is called an initial population, the length of which may be constant
or may vary from one generation to another. An example of gene {P2, P3, P4} is encoded as a
binary chromosome of length 8 and is shown in figure 5. The presence of a web page is coded as
1, otherwise as 0. Evidently, 2m different chromosomes are generated, where m is length of string.

Figure 5. Sample Binary chromosome

3.2 IOGM Evaluation function – Fitness function:
The evolution function quantifies the optimality of a solution so that a particular solution is
ranked against all the other solutions. The function depicts the nearness of a given solution to the
39

required outcome. It is an essential step in the overall process of genetic approach as it plays a
key role to evaluate survival capacity. The IOGM employs a robust fitness function which is
designed based on scores, ideal dimension and Euclidian distance of chromosomes.

In the Fitness function, Dc indicates number of genes in form of visiting patterns included in each
chromosome and Nc indicates the number of chromosomes.

Thus the population Np = Dc. Nc genes 1

The fitness function computed for each chromosome is expressed as a positive value and is to be
maximized. It is composed of three terms namely Term1, Term2 and Term3. The first term is the
sum of the length of the genes in chromosome C,

Term 1 : Tଵ ሺCሻ = ∑୮౟ ∈େ lengthሺG୧ ሻ 2

Where length(Gi) is the original length given to gene Gi. This term considers genes with only high
positive value in a chromosome. This drawback is balanced by considering the minimum support
in second term of the Fitness function. Let S be a minimum support.

୒୮
Term 2 : Tଶ ሺCሻ = ୟୠୱሺ|େ|ିୗሻ + 1 3

It reaches its maximum Np when the dimension of C is exactly equal to the minimum support S
and rapidly decreases when the number of genes contained in chromosome C is less than S.
The chromosomes that are present in the initial population are associated by the highest possible
variability values as far as the genes are concerned. The evaluation of the population alters
assigned values in process of creating new population. Moreover, the fact that genes belonging to
different chromosomes may not be guaranteed, as it depends on the number of genes Dc. For this
reason, fitness function third term is introduced, which measures directly the overall dissimilarity
of the genes in the chromosomes. Let D(Gi, Gj ) be the distance between gene Gi and Gj. is the
sum of the distance between the pairs of genes in chromosome C and the measures the total
variability expressed by C.

Term3: Tଷ ሺCሻ = ∑ୋ౟, ୋౠ∈େ,ୋ౟ஷୋౠ DሺG୧ , G୨ ሻ 4

The final form of the fitness function (FF) for chromosome C is derived from the equations 2, 3
and 4.
FFሺCሻ = α. Tଵ ሺCሻ + β. Tଶ ሺCሻ + γ. Tଷ ሺCሻ 5

Where α, β and γ are parameters that depend on the magnitude of the chromosome that represent
the genes. In particular α, β and γ are chosen so the contributions given by T1(C), T2(C) and
T3(C) are balanced. This approach is used by the evolution function of IOGM for every
chromosome C*, by means of the genetic operators.

FFሺC ∗ሻ = maxୡୀଵ,…..,୒ୡFFሺCሻ 6

3.3 IOGM Operators

The biologically inspired genetic operators are applied on population of chromosomes to generate
potentially new offspring. This is an iterative and fundamental step in genetic approach so as to
produce subsequent acceptable generation. The Selection, Crossover and Mutation are set of
operators designated by IOGM which transforms individual chromosomes stochastically. Each

40

chromosome has an associated value called fitness function that contributes in the generation of
new population by genetic operators. At each generation, the IOGM utilizes the fitness function
values to evaluate survival capacity of each chromosome. The IOGM operators create a new set
of population that tries to improve on the current fitness function values by using old ones.

Selection:
The selection process determines the number of times a particular individual chromosome is
chosen for reproduction from initial population as a mating pool for IOGM further operations.
The number of individual chromosomes obtain for the next generation is directly proportional to
its fitness value, there by mimic the natural selection procedure. The IOGM deploys the
stochastic universal selection to yield best offspring.

The key idea of stochastic universal selection gives preference to better individual chromosomes
by permit them to pass on their genes to the next generation and disallow the entry of worst fit
individual chromosome into the next generations. It principally works at the level of
chromosomes with no bias. The goodness of each individual chromosome depends on its
associated fitness function value.

The methodology of stochastic universal selection of IOGM functions to place the population NP
with equidistant markers on the wheel, which has as many slots as the population size NP. The
each NP individual chromosomes picks at random by spinning the wheel and a single random
number marker1 in the range [0, 1/Np] is generated. The Np individual chromosomes are then
selected by generating the Np markers, starting with marker1 and spaced by 1/Np, and decide the
individual chromosome whose fitness function value spans the location of the markers. The
function ET(i) is the Estimated Trials of individual chromosome i, LBET(i), UBET(i) are the
Lower and Upper Bound functions of ET(i) respectively. The individual chromosome achieves
the assured least number spins by choosing minimum times of LBET(i) and within the UBET(i).
This phenomenon continues until the number of individual chromosomes identical to the number
of markers that lie within the matching slot as shown in figure 6.

Figure 6. Stochastic universal selection

Crossover:

The crossover plays a vital role in the design and implementation of any robust genetic models.
The key focus of crossover creates new chromosome which is better than its parents, by taking
the best characteristics from each of the parents. Towards this, it exchanges the information
between randomly mated pair of fixed-length chromosomes by recombining parts of their mating
pool to produce new survival offspring. This operation, carry out probabilistically, combines parts
of two parent chromosomes to generate new offspring. The IOGM chooses single point crossover
technique as it generates successful new offspring.
41

The single point crossover technique of IOGM, initially pairs all the chromosomes at random in
the population obtained by stochastic universal selection procedure. The example pair of
chromosomes depicted with two different colours is as shown in figure 7(a).

Figure 7(a) Pair of chromosomes

Later, it selects the crossover point K randomly between 1 and L – 1, where L is the length of the
chromosome. This crossover point occurs between two bits and divides each individual
chromosome into two parts. The crossover point is represented with a thin vertical line across pair
of chromosomes as shown in figure 7(b).

Figure 7(b) single point crossover operation before crossover

Finally, it performs crossover operation on a pair of chromosomes at the crossover point. Then,
the parts of two parents after the crossover point are exchanged to form new offspring as shown
in figure 7(c). The single point crossover technique of IOGM repeats with a good number of trails
to arrive at a feasible offspring which push forward to the mutation process in obtaining the
optimal solution.

Figure 7(c) single point crossover operation after crossover
Mutation:

Mutation is a process which alters the genetic structure of the chromosome randomly. Its central
aim is to submit an application of genetic diversity from one generation of population of
42

chromosomes to the next. The optimal solution resides in a chromosome which is not presented in
the selection population. Thus, the previous genetic operators are unable to reach the global
optimum. In such situations, only mutation helps in generating new population with which
optimal solution can be attained. IOGM make use of bit by bit mutation to arrive at optimal
population.

The bit by bit mutation of IOGM alters every gene value in a chromosome from its initial
state with a low probability. It inverts the values of chosen gene. The value of 0 is replaced
with 1 and vice versa. This results as new chromosome with entirely new gene values. An
example of bit by bit mutation is shown in figure 8.

Figure 8. Process of bit-by-bit mutation
Mutation is an important part of IOGM, helps to prevent the population from stagnating at any
local solution of the solution space. With the new mutated gene values, the IOGM is capable to
arrive at better optimal solution than previously possible.

3.4 IOGM End function:
The IOGM results the best chromosomes as the process goes to the infinite number of iterations.
The process needs to be stopped after finite number of iterations. To take the decision on number
of iterations IOGM design an end function that stops the process at a finite value n, where n is the
stopping time of the process. The objective of end function determines the value of n, with which
the process of IOGM achieves optimal solution. Moreover, in stochastic process of IOGM, the
value must be probabilistic since no finite stopping time can assure the optimal solution always.
The methodology of the end function that works by considering population size Np, fitness
function value FF, length of gene Length(Gi) , crossover probability Pc and mutation probability
µ m . It interrupts IOGM process when no significant or sufficient improvement found in two or
more consecutive generations. In some cases, IOGM uses time threshold set by user, as a criteria
for ending the process.

43


3.5 IOGM Algorithm:

Step 01: Define the Evaluation function F.

Step 02: t ← 0 (iteration No =0, pop size =0)

Step 03: Initialize P (t).

Step 04: Evaluate P(t) (page from modeled data).

Step 05: Generate an offspring page O.

Step 06: t ← t+1 (new population).

Step 07: Select P (t) from P (t-1).

Step 08: Crossover P (t).

Step 09: Mutation P(t)

Step 09: Evaluate P (t).

Step 10: Go To 5 (while no termination (no of iterations)).

Step 11: End Condition P (t) Gives the output to the user.

The IOGM implemented with the theories and techniques of genetic approach for web usage
mining which derives optimal visiting patterns. In this process, encoding strategy along with
evolution function tuned the length of the chromosome and the population size of patterns
discovered by pattern discovery technique. The biologically inspired operators adjust the
probabilities of selection and crossover so that IOGM exhibits self learning capability to carryout
in producing survival patterns. The genetic diversity of mutation refines the patterns replacement
strategy. The termination criterion fixes the number of iterations that reduces the computational
cost of IOGM in deriving optimal patterns. These patterns help in investigating the web user
usage behaviour intelligently.

4. EXPERIMENTAL ANALYSIS
The proposed IOGM is experimented over a period of six months server side weblog data under
standard execution environment. For the IOGM algorithm number of visited patterns is given as
input to start the process. These patterns are generated by the pattern discovery technique.

A) The IOGM compared with the standard web mining technique in terms of performance. The
experimental results indicate that noticeable improvement of IOGM performance over the
standard web mining technique as shown in figure 9.

44


Figure 9. Comparison of performance of IOGM Vs standard web mining technique
.
B) The IOGM experimented for different cross over probabilities (Pc=0, 0.25, 0.5, 0.75, 1) on
different number of iterations. It shows high performance at the average mean Pc as shown in
. show
Figure 10.

Figure 10. Comparison of performance at different crossover probabilities

C) In addition, the standard analysis algorithms are applied on the collective output generated by
)
IOGM, the web user usage behaviour is identified as shown in Fig 11.

45


Figure 1 Web user usage behaviour
11.

5. CONCLUSIONS
The present model has proven the relevance of soft computing methodologies in the identification
of the desired patterns. The results are evident that the proposed Intelligent Optimal Genetic
Model has a promising future to arrive at optimal solution intelligently in the web usage mining.
web
The encoding strategy, evaluation function, fitness function, operators and end function of
proposed IOGM works together and achieve sustained adaptability to characterize web user usage
behaviour in unpredictable external environment. The nature of biological diversity of IOGM
e
prevents the population from stagnating at any local solution. Moreover, the stochastic process of
solution
IOGM, assures the optimal solution always. Hence the proposed IOGM helps to investigate user
M
usage behaviour on WWW in any application domain.
ACKNOWLEDGEMENTS
The authors record their acknowledgements to the authorities of Shri Vishnu Engineering College
for Women, Bhimavaram; Andhra University, Visakhapatnam and Acharya Nagarjuna
University, Guntur for their constant support and cooperation.

REFERENCES
[1] B. Minaei-Bidgoli, William F. Punch “Using Genetic Algorithms for Data Mining Optimization in an
Bidgoli,
Educational Web-based System”, http://www.lon-capa.org, 2003.
based
[2] B. Minaei-Bidgoli, G. Kortemeyer, and W. F. Punch. “Optimizing classification ensembles via a
Bidgoli,
genetic algorithm for a Web-based educational system”, In Proceedings of Joint International
Web based
Association for Pattern Recognition (IAPR) Workshops on Syntactical and Structural Pattern
Recognition (SSPR 2004) and Statistical Pattern Recognition (SPR 2004), 2004.
[3] C. Lopez-Pujalte, V. P. G. Bote, and F. de Moya Anegon., “Order-based fitness functions for genetic
Pujalte, “Order based
algorithms applied to relevance feedback”, Journal of the American Society for Information Science
and Technology, 54(2), pp: :152-160, 2003.
:152
[4] F. Picarougne, N. Monmarche, A. Oliver, and G. Venturini., “Web mining with a genetic based
algorithm”, In Eleventh International World Wide Web Conference, Honolulu, Hawaii, pp: 7
Honolulu, 7-11,
2002.
[5] Freitas, A.A., “A survey of Evolutionary Algorithms for Data Mining and Knowledge Discovery”,
See: www.pgia.pucpr.br/~alex/papers. A chapter of: A. Ghosh and S. Tsutsui. (Eds.) “Advances in
Evolutionary Computation”. Springer-Verlag, 2002.
Springer

46

[6] H. Kargupta. “The gene expression messy genetic algorithm”, In Proceedings of the IEEE
International Conference on Evolutionary Computation, pp: 631- 636, 1996.
[7] H. Kargupta and S. Bandyopadhyay. “A perspective on the foundation and evolution of the linkage
learning genetic algorithms”, The Journal of Computer Methods in Applied Mechanics and
Engineering, Special issue on Genetic Algorithms, 186, pp: 266-294, 2000.
[8] H. K. Tsai, J. M. Yang, and C. Y. Kao., “Applying genetic algorithms to finding the optimal Gene
order in displaying the microarray data”, In Proceedings of the Genetic and Evolutionary
Computation Conference (GECCO), pp: 610-617, 2002.
[9] H. K. Tsai, J. M. Yang, Y. F. Tsai, and C. Y. Kao., “An evolutionary approach for gene expression
patterns”, IEEE Transactions on Information Technology in Biomedicine, 8(2), pp: 69-78, 2004.
[10] Hyunchul Ahn, Kyoung-jae Kim, “Bankruptcy prediction modeling with hybrid case-based reasoning
and genetic algorithms approach, Applied Soft Computing, Volume 9, Issue 2, pp: 599–607, 2009.
[11] J. Han and K. C. Chang. “Data mining for Web intelligence”, IEEE Computer, pp: 54-60, November
2002.
[12] K. C. Wiese and E. Glen., “A permutation-based genetic algorithm for the RNA folding problem: A
critical look at selection strategies, crossover operators, and representation issues”, Biosystems, 72(1-
2), pp: 29-41, 2003.
[13] M. Martin-Bautista and M. A. Vila, “A survey of genetic feature selection in mining issues,” in Proc.
Congr. Evol. Comput. (CEC99), pp: 13–23, 1999.
[14] Mehmet Kaya, “Automated extraction of extended structured motifs using multi-objective genetic
algorithm” Expert Systems with Applications, Volume 37, Issue 3, pp: 2421-2426, 2010.
[15] M. F. Bramlette, “Initialization, mutation and selection methods in genetic algorithms for function
optimization” In Proceedings of the 4th International Conference on Genetic Algorithms, Morgan
Kaufmann, San Mateo, pp: 100-107, 1991.
[16] M. H. Marghny and A. F. Ali, “Web mining based on genetic algorithm” In Proceedings of ICGST
International Conference on Artificial Intelligence and Machine Learning, 2005.
[17] Mitsuo, G., Runwei, C., “Genetic algorithm & Engineering Optimization”, John Wiley & Sons 2000.
[18] Mu-Jung Huang, Hwa-Shan Huang and Mu-Yen Chen, “Constructing a personalized e-learning
system based on genetic algorithm and case-based reasoning approach”, Expert Systems with
Applications, Volume 33, Issue 3, pp: 551-564, 2007.
[19] Pei, M., Goodman, E.D., and Punch, W.F. "Pattern Discovery from Data Using Genetic Algorithms",
Proceeding of 1st Pacific-Asia Conference Knowledge Discovery & Data Mining (PAKDD-97),
1997.
[20] Romao, W., Freitas, A. and Gimenes, I. “Discovering interesting knowledge from a science and
technology database with a genetic algorithm”, Applied Soft Computing 4(2), pp: 121–137, 2004.
[21] Sankar K. Pal, Fellow, IEEE, Varun Talwar “Web Mining in Soft Computing Framework: Relevance,
State of the Art and Future Directions” IEEE transactions on neural networks, vol. 13, no. 5, 2002.
[22] S. Bandyopadhyay and S. K. Pal. “Pixel classification using variable string genetic algorithms with
chromosome differentiation”, IEEE Transactions on Geoscience and Remote Sensing, 39(2), pp: 303-
308, 2001.
[23] S. Bandyopadhyay, S. K. Pal, and B. Aruna. “Multi-objective GAs, quantitative indices and pattern
classification”, IEEE Transactions Systems, Man and Cybernetics - B, 34(5), pp: 2088-2099, 2004.
[24] S. J. Louis and J. McDonnell, “Learning with case injected genetic algorithms”, IEEE Transactions on
Evolutionary Computation, 8(4), pp: 3160-328, 2004.
[25] S. Hill, J. Newell, and C. O'Riordan, “Analysing the effects of combining fitness scaling and
inversion in genetic algorithms”. In Proceedings of the 16th IEEE International Conference on Tools
with Artificial Intelligence (ICTAI'04), pp: 380-387, 2004.
[26] Shital C. Shah, Andrew Kusiak, “Data mining and genetic algorithm based gene/SNP selection”,
Elsevier, Artificial Intelligence in Medicine 31, pp: 183-196, 2004.
[27] S. Ventura, C. Romero, A. Zafra, J. A. Delgado, and C. Herv´as, “A java framework for evolutionary
computation soft computing.” Soft Computing, vol. 4, no. 12, pp: 381–392, 2008.
[28] S. Y. Wang and K. Tai. “Structural topology design optimization using genetic algorithms with a bit-
array representation”, Computer Methods in Applied Mechanics and Engineering, 194, 3749-3770,
2005.
[29] S. Y. Wang, K. Tai, and M. Y. Wang. “An enhanced genetic algorithm for structural topology
optimization”, International Journal for Numerical Methods in Engineering, 65, pp: 18-44, 2006.

47

[30] Wang, X, A Abraham and KA Smith, “Soft computing paradigms for web access pattern analysis”. In
Proc. of the 1st International Conference on Fuzzy Systems and Knowledge Discovery, pp: 631 – 635,
2002.
[31] W. Fan, E. A. Fox, P. Pathak, and H. Wu, “The effects of fitness functions on genetic programming-
based ranking discovery for web search”, Journal of the American Society for Information Science
and Technology, 55(7), pp: 628 - 636, 2004.

Authors

Prof. V.V.R. Maheswara Rao received his Master of Computer Applications degree
from Osmania University, Hyderabad, India. He is working as Professor in the Dept of
Computer Applications at SVECW, Bhimavaram, AP, India. He is currently pursuing
his Ph.D. in Computer Science Engineering at Acharya Nagarjuna University, Guntur,
India. His Research interests include Webmining, Artificial Intelligence. He is member
of CSI, CI and ISTE.

Dr. V. Valli Kumari holds a PhD degree in Computer Science and Systems Engineering
from Andhra University Engineering College, Visakhapatnam and is presently working
as Professor in the same department. Her research interests include Security and privacy
issues in Data Engineering, Network Security and E-Commerce. She is a member of
IEEE and ACM.

48

AN INTELLIGENT OPTIMAL GENETIC MODEL TO INVESTIGATE THE USER USAGE BEHAVIOUR ON WORLD WIDE WEB

More Related Content

What's hot

Viewers also liked

Similar to AN INTELLIGENT OPTIMAL GENETIC MODEL TO INVESTIGATE THE USER USAGE BEHAVIOUR ON WORLD WIDE WEB

Recently uploaded

AN INTELLIGENT OPTIMAL GENETIC MODEL TO INVESTIGATE THE USER USAGE BEHAVIOUR ON WORLD WIDE WEB