Abstract Web pages along with Naive Bayes clustering


The distinctive webpage
recommendation for individuals is evident these days.

Web servers are
loaded with recommendation systems that analyse and recommend webpages for the
users. They use data implicitly obtained as a result of Web
browsing patterns of the users for recommending webpages.
The existing system
collects the Web logs and generates a cluster of
similar users and recommends pages to the user by actively analysing
it in online. However the time for analysing it in online is more. To optimize
this and increase the correctness of recommendation systems, a method
that applies Firefly based algorithm for recommending Web
pages along with Naive Bayes clustering is designed. User Web logs
are initially clustered in offline by using Naive Bayes clustering
technique. To find the similarity between the
active user queries with other users in the
cluster Firefly algorithm based similarity measure is used. The
proposed approach uses a probability based
clustering which eliminates the odd records while forming clusters.
Firefly algorithm meticulously searches the generated web
logs present in the cluster of the active user and recommends the top
pages. Firefly algorithm utilizes time efficiently, thus it is used for
processing in online. When pages are obtained, they are
ranked and the top pages that are more relevant to
the query are recommended.
The efficiency of the system can be evaluated
using measures like precision, recall-Score, Matthews’s correlation and
Fallout rate. The proposed approach is expected to improve time
utilization in online process as well as recommends
more accurate Webpages.   


Introduction- Web
page recommendation system is a sub-domain of recommendation systems that
recommends a set of Web pages to the users based on their past browsing
patterns. It is done by applying special mining techniques on the data that are
previously gathered from the users which in turn discovers and extract
information from Web documents and services. The major concern is to find
reliable and efficient recommendation algorithms. Recommendation system
typically produces the result by following one of the two ways – through
collaborative and content based filtering.


A.    Collaborative


recommendation system has wide use of collaborative filtering for recommending
items. This method lies on collecting and processing the information’s on
user’s behaviours or activities and then predicting the items relating to their
similarity with other users. Collaborative filtering approach builds a
structure from the users past behaviours and decisions of other similar users.
This model is used to predict user interested items. Since collaborative
filtering is independent of machine analysable contents, it is capable of
recommending for complex items accurately without “understanding” of the item


 B.    Content
Based Filtering 

Content based
filtering is a widely used approach for designing recommendation systems. This
technique is lies on a definition of item along with a user’s preferred
profile. In a content based recommendation systems, the keywords are considered
as user’s interest. It utilize a series of distinct property of an item for
obtaining and recommending items with same properties. These approaches are
continually combined as Hybrid Recommendation Systems. These algorithms try to
recommend items based on examining the items that are liked by a user in the
past or in the present. In general, various items of candidate set are compared
with items that are rated by the user in the past and the best matching items
are recommended.



Literature survey

system has a major role of recommending personalized items for the users
based on their interest in a web services. The web
also contains a rich and dynamic information’s. The amount
of information on the web is growing dynamically,
as well as the number of web sites and webpages per web site. Predicting the behaviours
and needs of a web user has gained importance. Many webpage
recommendation system were developed in the
past, since they compute recommendations in
online process, their time utilization should
be efficient. A system 4 that uses support vector
machine (SVM) learning based model was
developed for computing similarity between two items
which performed better than latent
factor approach for group recommendations. Since the
matrix representation was followed, the
data sparsity problem was solved.
However, the system was not able
to stably scale when size of the group
dynamically increased.


recommender systems which combines more number of recommendation techniques was designed
5. It eliminates any weakness which exist when only
one recommender system is used. There are several ways in which the
systems can be combined, such as weighted hybrid
recommender in which the score
of a recommended item is computed from the results
of all of the available recommendation techniques present
in the system. However, data sparseness was still a problem, the system may
generate week recommendations if few users have rated the
same items and also the system doesn’t
overcome the cold start problem. 


sensors can acquire hundreds of contiguous bands over a wide
electromagnetic spectrum for each pixel. To reduce computational
cost and eliminate an actual classifier within the band searching process, an
improved firefly algorithm based band selection method 8 was used. The
Firefly algorithm is an evolutionary optimization algorithm proposed
by Yang 13. After the initializations of parameters, the brightness is calculated
with the objective function .Then the moment states were
evaluated and the bands are selected. Firefly algorithm also had
a faster convergence even at the size of the
data is larger.



Further, to improve
the accuracy of similarity measure, firefly algorithm based similarity measures
are also introduced 10.It considered separate effects for ratings of
users with similar opinions and conflicting opinions. In order
to generate initial population of fireflies, half of population randomly
generated and the other half of population are randomly generated. Mean
absolute error was chosen as objective function to measure recommendation accuracy which
is obtained by difference between predicted rating and real rating.


An optimal
similarity measure via a simple linear combination of values and ratio of
ratings for user-based collaborative filtering provides better results. It
increased speed of finding nearest neighbours of active user and reduce
its computation time. Similarity function equation based on Firefly algorithm
was simpler than the equation used in traditional metrics
therefore, the proposed method provided recommendations faster than traditional


Graph colouring problems are
generally discrete. Algorithms to discrete problems are
quite complex. A new algorithm based on Similarity and
discretize firefly algorithm directly without any other hybrid
algorithm was developed 11. It was adoptable to dynamic graph

A system for
assigning an electronic document to one or more predefined categories
or classes based on its textual context and use of agglomerative
clustering algorithm was developed 6. This type of
clustering along with sample correlation coefficient as
similarity measure, allowed high indexing term space reduction factor with
a gain of higher classification accuracy.


In order to
minimize noise and outlier data, a modified DBSCALE algorithm using Naïve Bayes
has been designed 7. This algorithm is basically a prospect based
utility. This function is used to
estimate the outlier cluster
data and increase the correctness rate of algorithm on given
threshold value. Since Naïve Bayes is a probability based function,
it removes outlier cluster data and increases the correctness rate according to
threshold value. It also computes maximum posterior hypothesis for outlier
data. In order to minimize noise and outlier data, a modified DBSCALE algorithm
using Naïve Bayes has been designed 7. This algorithm is basically a prospect
based utility.


This function is
used to increase the correctness rate of algorithm on given
threshold value and to estimate the
outlier cluster data. Since Naïve Bayes is
a probability based function, it removes outlier cluster data and
increases the correctness rate according to
threshold value. It also computes maximum posterior
hypothesis for outlier data. The memory
based collaborative system uses matrix
based computation and solves data sparsity problem but, scalability
of the system cannot be stable when size of the group dynamically increases.
Hybrid system could be helpful in overcoming
the scalability issue but it again leads to cold start problem.


eliminate outliers as well as overcoming other two problems Naive Bayes clustering,
a probability based method was used in past.
Firefly algorithm has a faster convergence and searches all
possible subsets with better time utilization. Thus, to design an efficient
recommendation system, Naïve Bayes method can be followed for
clustering in offline. Since the time complexity should be
less, Firefly algorithm that is more
efficient in terms of time utilization, it can be used for
calculating similarity in online. Combination of these two technique
might increase the accuracy of the
recommendation system as well as results in efficient
time utilization.                





 III.   Overview of the proposed work 


Initially, the web log files are obtained from
the 1 America Online Inc. The log files consists of five
fields i.e. anonymous ID for individual user, query of each user along
with query time, list of URLs which user proceeded and its
rank in the result. These logs are collected
and grouped based on anonymous ID. The URL among all
the users are obtained and its content are downloaded and
processed. The processing of data includes removal of
stop words from the URL’s data and
keyword extraction. Similar users are clustered based on fetched
keywords by using Naïve Bayes clustering technique which provides efficient
clusters compared to clustering by the use of association rules. The created
clusters are given to online component. In online process, when an active user
gives a query, the keywords from the query is extracted. The
similarity between the extracted keywords with the other users
in the same cluster of the active user
is calculated using Firefly similarity measure. The
similarity values are sorted along with the web pages
browsed by similar users in the cluster. The top k web pages are
recommended for the active user
as a result.            




IV. The proposed


The proposed
system follows a linear process of initially collecting the
web logs and processing them followed by clustering similar users
by Naïve Bayes clustering technique and finally generating
recommendations based on a similarity measure from firefly


Pre-processing of Web Logs


web logs are collected form 1 AOL Inc. It consists of 20
million web queries from 650 thousand real users over 3
months. The data set includes anonymous ID, query, query
time, item rank and click URL. The log file contains
many number of users along with the web pages visited by
them. It is validated and separated based on anonymous ID. The user
is separated into individual file using anonymous ID. The content from
the URL are fetched and downloaded.
Those keywords are processed which undergoes stop
words removal and
stemming process. The final keywords are then
extracted. The features like keywords, Timings, Frequency, Click URL and
Revisit are fetched. The user profile is constructed using those
features. The user profile that constructed is based
on the features that are taken
form the user log files.


Timing: The timing
that the user spent on that particular URL

Frequency: The
amount of time the user visited the URL

Clickstream: The
number of click stream that are visited by user

Whether the user visited the web page


The keywords are
generated from the data fetched form the
URL. Timing for each URL is estimated from
the given date and time by calculating the difference
between the each URL that are searched in a single
day by having some time constraints. Frequency
is hence calculated such that number of times the user
clicked the URL. The clickstreams are those that are
clicked by the user for additional information. The timing
of revisit is calculated such that to decide whether the
user preferred it much or not. Keywords:
Keywords are those which are extracted from the URL.
The information from the URL is hence collected and processed to
obtain features of the user.  



Naïve Bayes Clustering 


Clustering, also
known as unsupervised classification, is a descriptive task with many
applications. Clustering is decomposition or partition of a data set into
groups such that the object in one group are similar to
each other but as different as possible from the
object in other groups. Three main approach for clustering of data is partition
based clustering, hierarchical clustering and probabilistic model
based clustering. Probabilistic model based clustering is a
soft clustering were an object can be in many cluster
following a probability distribution. A clustering is useful if it produces
some interesting insight in the problem that we
are analysing. Naïve Bayes clustering is also a probabilistic clustering technique
that is based in Bayes theorem with strong independent
assumption between features. The feature variables can
be discrete or continuous. This probabilistic clustering lies on nominal and
numeric variables in the data set and its novelty lies in the use of mixture of
truncated exponential (MTE) densities to model the numeric variables. In Naïve
Bayes clustering the class is the only root variable and all
the attributes are conditionally independent given the class. The
clustering problem reduces to take a data set of instances
and a previously specified number of clusters (k), and work out
each cluster’s distribution and the population distribution between
the clusters. To obtain these parameters the expectation maximization (EM)
algorithm is used. Since Naïve Bayes clustering is
a probability based techniques. The items belongs to the
cluster if and only if it has a relation to it. This helps in
eliminating outlier data in the process of clustering. It also provides proper
clustering with less computations. The given dataset is divided into two parts,
one for the training and other for testing. For each
record in the test and train databases, the distribution of the class
variable is computed. According to the obtained distribution, a value for the
class variable is simulated and inserted in the corresponding cluster. The
log-likelihood of the new model is computed. If it is higher than the initial
model, the process is repeated. Otherwise, the process is stopped,
obtained clusters are returned.  


Optimisation Using Firefly Algorithm


algorithm is an evolutionary algorithm that is based on the
behaviour of fireflies. Fireflies live in colonies and cooperate for the
survival of the colony. Generally, in order to model the behaviour of
fireflies, three assumptions will always be considered i.e. all fireflies are
homogeneous, Attractiveness of each firefly is related to its level of
brightness, rightness of firefly is determined with an exponential
objective function. Each firefly always emits a kind
of light that by which attracts other fireflies. The amount of accessed
light depends on parameters such as distance and absorption coefficient of the
surroundings. The longer the distance the lesser the amount of accessed light
will be. Also in surroundings with high light absorption coefficient such as
foggy weathers, the intensity of light decreases. The
certain issue is that every firefly regardless of its gender has
always been attracted to and moved toward the brighter firefly.
Firefly has a light intensity of its own. The key concept is, the firefly with
low light intensity is always attracted to the firefly with high light
intensity. This concept can be incorporated for calculating similarity. By
using firefly based similarity measure unique and distinguished results can be
obtained which is a useful feature for ranking. It can deal with highly non-
linear, multi-modal optimization problems naturally and
efficiently. It does not use velocities, and there
is no problem as that associated with velocity in PSO. The
speed of convergence is very high in probability of finding the global
optimized answer. It has the flexibility of integration with other optimization
techniques to form hybrid tools. It does not require a
good initial solution to start its
iteration process. Each web pages visited by
the user i are considered a firefly. The number of user visited the
particular page is assumed as the light intensity of the firefly. The objective
function is formulated based on the frequency and duration. Frequency is
calculated as the ratio to the number of visits per page to the average vests
of all pages.



The duration is
the ratio of duration of page to the total duration of all the pages visited by
the user. Thus, the objective function can be defined as in equation 5.1
Interest (i)= 2*Frequency (i)*Duration (i) Frequency (i)+Duration (i) (5.1)
  The interest of all users in the cluster is calculated. Then the pages
to be recommended are found by using page rank algorithm 2 on the obtained
result. The results after applying page rank algorithm is given as the
recommended web page to the user.    


D. Ranking the Web


The result, set of
web pages obtain should ranked in an order that the user might have higher
interest. Thus, they are
ranked in a sorted order based
on the interest of the active user. The association
rule checks the maximum possible combinations
which provides more accurate pages.



Recommendation Process


The URL that are
to be recommended will be identified based on ranking and similarity measure.
The similarity measure is calculated among the users by comparing their similar
interest. From the obtained result of pages, page rank algorithm
is used to rank the most relevant pages to the user. Thus, resultant URL’s are
recommended to the users. Hence
the web page that is to be recommended to
the user will be more relevant. The use of Naive Bayes clustering
will eliminate the outliers and Firefly based similarity calculation will
check all the subsets of the clusters.