Enhancing proxy based web caching system using clustering based pre fetching with machine learning technique

IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
__________________________________________________________________________________________
Volume: 03 Special Issue: 07 | May-2014, Available @ http://www.ijret.org 463
ENHANCING PROXY BASED WEB CACHING SYSTEM USING
CLUSTERING BASED PRE-FETCHING WITH MACHINE LEARNING
TECHNIQUE
V.Sathiyamoorthi1
, P.Ramya2
1
Assistant Professor, Department of Computer Science, Sona College of Technology, Sona Nagar, Salem, Tamilnadu.
2
PG Scholar, Department of Computer Science, Sona College of Technology, Sona Nagar, Salem, Tamilnadu.
Abstract
Data Mining is a process of extracting knowledge from various data source. Web usage mining is the area of data mining which deals
with the discovery and analysis of usage patterns from web data, specifically web logs, in order to improve web based applications.
Considering these, Web caching is a Data mining technique which is used to reduce user perceived latency when user is accessing the
web pages. Web pre-fetching is the scheme where web pages are pre fetched into the intermediate server (proxy) cache before user
accessing it. These two techniques can complement each other since web caching exploits the temporal locality whereas web pre-
fetching utilizes the spatial Locality of web objects. To improve the performance of web access, web caching and pre-fetching
techniques are being integrated using clustering based pre-fetching algorithm. In this paper, pre-fetching using clustering technique is
combined with SVM (Support Vector Machine)-LFU algorithm, a machine learning technique for web proxy caching. By analysis it is
shown that SVM technique is better than clustering based pre-fetching technique using caching policy like LFU considering
bandwidth utilization and access latency.
Keywords: Web Caching, Web Pre-Fetching, Clustering, SVM Machine Learning Technique.
------------------------------------------------------------------------***------------------------------------------------------------------
1. INTRODUCTION
In today’s world, the speed of distribution of information from
one place to another is of vital importance. Web pages which
are quite slow to download may create a lack of user’s interest
in accessing them. Hence to retain the viewership, we have to
ensure that the pages are accessed quite fast. The traditional
cache replacement techniques used often fail to increase the
cache hit ratio. In this project different mechanisms are being
proposed for improving the web cache performance. One such
technique is integration of web pre-fetching and caching to
anticipate the user’s requests and also the clustering technique
is used so that the pre-fetching engine performs faster and
responds to the user’s requests immediately with most related
web objects. As web caching, pre-fetching and clustering
technique plays a vital role here we discuss in detail on all
data mining techniques. After the clustering process, it is
combined with the SVM-LFU algorithm in order to show that
the performance improvement using SVM would be
comparatively better than implementing with clustering based
pre fetching technique. The performance is measured using the
performance metrics as HR (hit ratio) and BHR (byte hit
ratio).In section 2 discussion is based with the techniques as
Web Caching and Pre Fetching, in section 3 Clustering based
Pre Fetching technique is discussed, section 4 deals with the
Classifier technique i.e Support Vector Machine (SVM) and
the combination of SVM and Cluster is discussed and final
section deals with the performance of combining the SVM
with Cluster is discussed.
2. WEB CACHING
Web caching is a technique which is used for caching as many
web pages in the cache to improve network performance.web
caching can be done either at the client side, or at the proxy
server or at server side. Web caching aims to reduce network
traffic, server load, and the user perceived retrieval latency.
The main component of a caching system is its page
replacement policy, which needs to make good replacement
decisions when its cache is full and a new document arrived.
In this work, web usage mining is used to optimize the
existing web cache policy for better performance.
2.1. Caching Strategies
Caching strategies decide which document to remove from the
cache when space is needed to new ones. Their goal is to make
the usage of the cache as efficient as possible, by trying to
optimize one or several of the performance criteria of caching
cache hit rate, byte hit rate, or response time. Additionally the
strategies themselves should work efficiently, i.e CPU and
memory usage should also be as low as possible. To fully
optimize the cache usage, all future requests must be known
which impossible. Therefore the strategies use heuristics to

__________________________________________________________________________________________
decide which documents to replace and which to keep. Few
important factors that influence the replacement strategies are:
 Recency: time of the last reference to the object.
 Frequency: number of request to an object
 Size: size of the web object
 Cost of retrieval: cost to fetch an object from original
server.
When these factors are taken into account it includes benefits
as, reducing the cost of connecting to the internet. Reducing
the Latency of today’s WWW.
Some of the standard strategies used are,
LRU: Least Recently used (LRU) is the most common
caching strategy. The items not used for the longest time are
removed from the cache to make way for new ones. This
algorithm leads to a high rate while causing little overhead.
LFU: Least frequently used (LFU) is the caching strategy
where the items not used for long time are removed from the
cache to make way for new ones. This algorithm lead to
maximum number of hit rate compared to other caching
techniques.
FIFO: First in first out (FIFO) is a caching technique which is
used similar to stack operation. The item in the first entry will
be removed first when new one comes if the cache memory is
filled. This algorithm gives the average number of hit rate.
Apart from these some of the intelligent web caching
algorithms are Intelligent Client Side Web Caching Scheme
based on Least Recently used algorithm and Neuro-fuzzy
system by Ali &Shamsuddin (2009)],NNPCR by cob &
Elaarag (2008) and many more algorithms. In case of Least
Recently used algorithm the important factor called the
recency factor is being neglected.
The following figure shows the exact loction of proxy cache
and the proxy server.proxy cache is the one which replaces the
web pages according to the requests by the users whereas the
proxy server serves the web page that is requested by the
user,immediately if it is present in the server else it is
forwarded to the cache and accordingly the replacement is
being done.
2.2. Performance Measures
There are standard metrics to analyze the efficiency and the
performance of web caching techniques, they are Hit
ratio(HR),Byte Hit Ratio(BHR) and Latency Saving Ratio
(LSR) are the most widely used metrics used in evaluating the
performance of web caching.
HR = /N (1)
BHR = / (2)
LSR = / (3)
Where N is the total number of requests, ti is the time taken for
retrieval of object, and bi is the size of the ith request.
2.3 Web Pre-Fetching
Web pre-fetching is a mechanism which is used for pre-
fetching web pages into the cache before the actual request
arrives. When combining web pre-fetching and web caching
they can complement each other since the web caching
technique make use of temporal locality, whereas web pre-
fetching technique exploit the spatial locality of web objects.
The problem is to determine which page has to be pre-fetched
and cached. This problem is aggravated by the fact that there
are wide spectrums of users and each has their own
preferences. This work tries to solve the above problem by
grouping the users based on their access patterns. Due to
grouping of user’s, we avoid the task of going into the
individual preferences which will be quite a formidable task.
The work presented here will integrate web caching and
clustering based web pre-fetching scheme to improve the
performance of proxy cache. Web pre-fetching scheme can be
classified into two types: short term pre-fetching and long
term pre-fetching schemes. A brief summary of the two
approaches are as follows, they are short term pre fetching
scheme and long term pre fetching scheme.
2.3.1. Short Term Pre-Fetching Scheme
The short term pre-fetching scheme pre-fetches web pages in
the cache by analyzing the web cache’s recent access history.
Based on the analysis, the scheme computes the cluster of
closely related web pages and pre-fetches clusters of web
pages from the origin web servers (Chen, Qui, Chen, Nguyen
& Katz, 2003).
2.3.2. Long Term Pre-Fetching Scheme
In Long term pre-fetching scheme, the popular web pages are
identified by analyzing the global access pattern for the web
page (Lee, An, & Kim, 2009) in this scheme, objects with high
access frequencies and without longer update time intervals
are more likely to be pre-fetched. Thus, web pre-fetching is a

__________________________________________________________________________________________
pro active approach where the web pages are pre-fetched into
the proxy server cache from the origin web servers.
2.3.3. Measuring the Performance of Pre-Fetching
To judge the success of a pre fetching system and to tune the
parameters used, the performance of the system must be
measured. The following criteria can be used to do this
 Usefulness of Predictions/Pre-fetches: the percentage
of fetched pages that had already been predicted is pre-
fetched.
 Accuracy of Predictions/Pre-fetches: The percentage
of predicted or pre-fetched pages that were later
actually requested by the user.
 Practical accuracy of predictions: the probability that
one of the received predictions was correct.
 Coverage: the percentage of actual fetches which were
preceded by the predictions.
 Network traffic Increase: The volume of network
traffic with pre-fetching enabled / the volume of traffic
without pre-fetching.
These are some of the measures used to detect the
performance of pre-fetching technique. Most of the existing
pre-fetching techniques will employ single object pre-fetching
technique, which can be handled by traditional cache
replacement algorithms. But when clustering based pre-
fetching technique is used, in which multiple objects are pre-
fetched, so which cannot be handled by traditional algorithms.
So the proposed cache replacement algorithm will consider
this issue by modifying cache replacement algorithm and
provide better performance.
3. CLUSTERING BASED PRE-FETCHING
TECHNIQUE
As we had seen the web caching and pre-fetching process now
we move into the process of integrating the web caching and
pre-fetching technique using the data mining based clustering
pre-fetching technique. Specifically to move into the
integration process let us see the clustering Process.
3.1. Clustering Technique
Clustering is a process of dividing the given data into group of
objects. It is an unsupervised learning Technique of hidden
data in data mining. Each group can be clustered together if
the objects are similar among themselves and dissimilar to
objects of other groups. Clustering will assign the points to
finite system of K subsets which is called a cluster. From
machine learning perspective clusters correspond to patterns
and these patterns are hidden. A computational requirement
which deals with large database is known as cluster analysis.
In this paper the clustering based pre-fetching scheme is being
used in which the Web navigational graph (WNG) is
constructed for each user independently using a user’s session
time interval. At the end of each time interval new
navigational graph is constructed for each user based on
content of log files. Each node in the WNG represents a web
object requested by each user and each edge represents user’s
transition from one web object to another and a weight is
assigned to each edge which represents the number of
transitions between those objects.
The clustering algorithm gets the content of WNG as input.
Support and confidence are the parameters used to keep track
of frequently visited pages by user. A threshold value is fixed
for those parameters and those edges which have values less
than this threshold is removed. Support is defined as the
frequency of navigation between two nodes u1 and u2. The
confidence is defined as freq (u1, u2)/pop (u1) where pop (u1)
is the popularity of u1.Popularity of the node is defined as the
number of incoming edges into that node. The WNG is
partitioned in to sub graphs by removing those edges having
low support and confidence values. The nodes in each
connected sub-graph are identified as a cluster. When a user
requests any one of the nodes (Web objects) in a cluster and if
it is found in the long- term cache or in the short-term cache,
all the remaining nodes in that cluster can be pre-fetched in to
the short-term cache by predicting that those objects will be
requested by the user in the near future [20]. This pre-fetching
is done during the browser idle time and this pre-fetching
helps in reducing the user perceived latency time. If the access
count of a Web object is greater than the threshold value then
that Web object is moved from the short term cache to long
term cache. LFU technique is used for removal of pages from
the short term cache if sufficient space is not available for
caching a new Web object.
3.1.1. Preprocessing Step
A log file generated by the proxy server consists of the
following fields,
 Machine IP address making the request
 Cache hit/ Cache miss and Response code
 Size of the requested object measured in bytes.
 URL of the requested page
 Information if the request is redirected to other server.

__________________________________________________________________________________________
Fig.2.Architectural diagram of proposed work
3.1.2. Web Navigation Graph (WNG)
A weighted directed Web graph (𝑥,𝑦) is used to represent the
requests of each user, where each node 𝑥 represents a Web
object and each edge 𝑦 represents a user’s transition from one
Web object to another. The weight of each edge represents the
number of transitions in the set. To make the size of the WNG
manageable, the edges are removed whose connectivity
between two Web objects is lower than a specified threshold.
Support and confidence are the two parameters that determine
the connectivity between two objects [1].
Let : <𝑥𝑖,𝑥𝑗> be an edge from node 𝑥𝑖R to node 𝑥𝑗R. Support
of 𝐺, denoted by 𝑓𝑟𝑒𝑞(𝑥𝑖,𝑥𝑗) is defined as the frequency of
navigation steps between𝑥𝑖𝑡𝑜𝑥𝑗R. Confidence of g is defined
as 𝑓𝑟𝑒𝑞(𝑥𝑖,𝑥𝑗)/𝑝𝑜𝑝(𝑥𝑖), where 𝑝𝑜𝑝(𝑥𝑖) is the popularity of𝑥𝑖R.
By this definition, the support value of the edge (𝑥3,𝑥4) for the
user X in Figure 3 is(𝑥3,𝑥4) = 1 and confidence value is
𝑓𝑟𝑒𝑞(𝑥3,𝑥4)/𝑝𝑜𝑝(𝑥3) = 0.5. If the support threshold chosen is
very less, too many less important user’s
Transitions for clustering may be included and if the chosen
threshold value is high, many interesting Transitions that
occur at low levels of support may be missed.
Fig.3.Web Navigational Graph (WNG)
3.1.3. Web Clustering Algorithm
The algorithm for clustering inter-site Web pages is described
below [20]. A weighted directed Web graph (𝑥,𝑦) that
represents the access patterns of a user is used. This graph is
partitioned into sub graphs by filtering edges with low support
and confidence values. The nodes in each connected sub graph
in the remaining navigational graph will form a cluster. The
inputs to this clustering algorithm will be Web navigational
graph, the number of users, support threshold and confidence
threshold. All the edges with support or confidence value less
than the corresponding threshold values are removed. BFS
(Breadth First Search) algorithm is applied to the navigational
graph. BFS takes a node in the graph (called as source) and
visits each node reachable from the source by traversing the
edges. It outputs a sub-graph that consists of the nodes
reachable from the source. This procedure is applied for all the
nodes of the graph. All the nodes in each connected sub-graph
forms a cluster. The time complexity of BFS is (|𝑥| + |𝑦|)
where |𝑥|the number of nodes is and |𝑦| is the number of edges
in the graph [1].
Fig: 4 BFS Algorithm
3.2. Clustering Based Pre-Fetching
The following are the steps that take place in the proposed pre-
fetching method
• A user requests a web object.
• Using the IP address, the proxy identifies the user and
maps the user to a particular user group. Given that the
Clusters of Web objects are known, the proxy searches
inside the existing clusters of that user group to find in
which cluster the requested object exists.
• All the remaining objects from the selected cluster are
Pre-fetched from the origin server by the proxy and
they are loaded into the short-term cache during the
browser idle time.
• The proxy sends to the user his/her requested object.
Input: G (u, v);
Unmark all nodes u;
Choose some starting
node x;
Mark x;
List L = X;
Sub graph T= x;
While L nonempty;
{Choose some node y
from front list;
Visit y ;}
For each unmarked
neighbor w
{Mark w;
Add it to end of list;
Add edge yw to T ;}
Client group 1
U1, u2, u3, u2, u4
U2
U4
U1
U3

__________________________________________________________________________________________
The following describes in detail about the BFS
algorithm with support and confidence value.
Fig.5 Clust Web Algorithm
4. PROPOSED WORK COMBINING
CLUSTERING WITH SVM-LFU
4.1. Intelligent Web Proxy Caching Algorithms
Compared to traditional caching approaches, intelligent Web
caching methods are more efficient.
Details about intelligent caching methods are found in [4] and
that of conventional replacement policies are found in [23]. In
our work, the proxy cache is divided in to short-term cache
and long-term cache. The Web objects requested for the first
time are loaded in to short-term cache. LFU algorithm is used
for managing short-term cache. Those objects visited more
than once from the short- term cache are moved to long-term
cache. In case of BPNN and NNPCR caching technique, the
recency factor which is considered as an important factor is
being ignored in these techniques. Moreover the LRU caching
technique when used ignores the frequency factor during the
replacement of cache content. From the above studies it is
observed that intelligent caching technique can be employed
either individually or can be combined with LFU technique.
4.2. Support Vector Machine (SVM)
SVM (Support Vector Machine) is a technique useful for data
classification. SVMs are supervised learning models
associated with learning algorithms for analyzing data and
recognizing patterns which are used for classification and
regression analysis. It takes a set of input data and predicts to
which one of the two possible classes, the given input data
belongs to. Given a set of training datasets, each marked as
belonging to one of two classes, a SVM training algorithm
constructs a model that classifies new data inputs into one
category or the other. SVM determine the hyper-plane to do
binary division that is used to find the linear boundary
between two classes such as Positive (Class1) and Negative
(Class0) [5]. The hyper-plane is placed in between two classes
and it is oriented like, the distance between the plane and data
point in each class is maximized. The nearest data points are
called as Support Vectors. In an arbitrary dimensional space a
separating hyper plane can be written,
𝑤.𝑥 + 𝑏=0
Where, 𝑏 is bias, 𝑤 the weights and𝑥is the input vector.
The decision function can be written as,
𝐷(𝑥) = 𝑠𝑖𝑔𝑛 (𝑤.𝑥 + 𝑏)
If the sign of the decision function is positive the object is
classified as Class 1 (Positive Class), otherwise the object is
classified as Class 0 (Negative Class)
𝑤.𝑥 + 𝑏=1 (𝐶𝑙𝑎𝑠𝑠 1)
𝑤.𝑥 + 𝑏=−1 (𝐶𝑙𝑎𝑠𝑠 0)
Fig.6 SVM
SVM classifier is used to classify the Web objects as class 0 or
class1, while moving them from short term cache to long term
cache. When a user requests for a Web object, simultaneous
search is made in short term cache as well as in the long term
Input Parameters:
Confidence,support,
G(i)=navigation graph,
K=number of client groups:
Begin
For i=1 to K
{ CutWithConfidence(G(i),Confidence);
CutWithSupport(G(i),Support);
TraverseWithBFS(G(i));
}
End
Procedure
CutwithConfidence(G(i),Confidence)
{
For j=1 to num-of-edges
If (edge[j] <Confidence) remove edge[j];
}
Procedure CutwithSupport(G(i), Support)
{
For j=1 to num-of-edges
If (edge[j] <Support) remove edge[j];
}
Procedure Traverse With BFS (G(i))
{
int c1[i]=0;
Repeat
{C[c1[i]]=BFS(G(i));// C; a list array of
traversed nodes
C1[i]++;
} until (all nodes of G(i) are traversed);
}

__________________________________________________________________________________________
cache. If the requested Web object is available in the short-
term cache, its access count is increased by one. If the access
count is greater than the threshold value then that web object
is given as input to SVM classifier for classification. If it is
classified as class 0, it is moved to the bottom of long-term
cache. If it is classified as class 1, it is moved to the top of
long-term cache. If there is no space in the long-term cache,
the Web object(s) placed at the bottom of the long-term cache
are removed to provide space for this new Web object.
Meanwhile a copy of this web object is given to the requested
user. If that Web object is found in the long-term cache,
classification of that Web object is done again. If it is re-
classified as class 1, it is moved to the top of the cache (long-
term cache) else it is moved to the bottom of that cache. It is
then given to the requested user and pre-fetching of other web
objects if any belonging to that cluster in to the short-term
cache is initiated If cache miss occurs, the requested object is
searched in all the clusters generated by Clustering algorithm
for that user. If it is found in any one of those clusters, that
Web object is fetched from the original server and it is sent to
the user as well as placed in to the short term cache. Other
Web objects that are present in that cluster will be pre-fetched
during browser idle time and will be placed in the short-term
cache. If the requested object is not found in any of the
clusters of that user, then the required Web page is fetched
from the original server. A copy of it is placed in the short
term cache and it is sent to the requested user. Thus by
combining Web caching and pre-fetching it is possible to
increase the hit ratio, decrease user perceived latency and
reduce the origin server load.
5. PERFORMANCE EVALUATION
5.1. Dataset
The scheme explained in this paper is tested with a dataset.
The dataset is obtained from a proxy server installation
www.ircache.net. The raw proxy server log file for the data set
contained the details of more than 1 million requests. The log
cleaning process was applied on the dataset, and moreover the
dataset contained about 610,634 entries fir for analysis. 70%
of all the requests have been used for the user’s access pattern
analysis, creating training dataset and testing.[1].The
remaining 30% will be used for the testing the scheme.
5.2. Performance Metrics
 Hit Ratio(HR)
HR is the percentage of the total number of requests
served by the cache over the total number of requests
given.
 Byte Hit Ratio(BHR)
BHR is the percentage of the number of bytes
corresponding to the requests served by the cache over
the total number of bytes requested.
Generally HR and BHR are calculated using different values
of support and confidence using SVM and LFU pre-fetching.
The performances for both of these methods for various cache
sizes are given in the table below. Moreover increase in the
support and confidence value leads to increase in number of
web objects in the created clusters. users experience decreased
access latency while using SVM pre-fetching because of
increase in percentage of cache hits. this ensures lesser load on
the original server.
6. CONCLUSIONS AND FUTURE WORK
In this paper the web access is improved by preprocessing the
log file and constructing web navigational graph. Moreover
clustering of Web objects in the WNG is done using a
clustering algorithm. Frequency of frequently used Web
objects is monitored by Support and Confidence values. The
pre-fetching of requested web objects is done during the idle
time either from short term cache or long term cache. Using
access count a Web object in short-term cache is moved to
long-term cache after classification using SVM algorithm. A
cache miss in both the caches leads to fetching of that object
from the original server and a copy of it being placed in the
short-term cache. The SVM-LFU algorithm is compared with
that of clustering based pre-fetching technique and the
performance would be shown to better in SVM than clustering
based pre-fetching.
This work can be further enhanced with the content
distribution network to improve the performance of the server
overall using some of the web caching and pre-fetching
technique integrated.
REFERENCES
[1] G.Pallis, A.Vakali, and J.Pokorny,”A Clustering-based
Pre-fetching Scheme on a Web Cache Environment”.
Computers and Electrical engineering,
34(4),(2008).pp.309-323.
[2] Ali W, Shamsuddin S M, Ismail A S, “Intelligent Web
Proxy Caching Approaches Based on Machine
Learning Techniques”. Decision Support Systems 53
(2012) 565-579.
[3] W.Ali, S.M.Shamsuddin, A.s.Ismail,”A Survey of Web
Caching and Pre fetching”. International journal of
advances in soft computing and its application (2011).
[4] S.Podlipnig, L.Boszomenyi,”A Survey of Web Cache
Replacement Strategies”,CM computing surveys
35(2003) 374-398.
[5] C.W.Hsu, C.C.Chang, C.J.Lin,”A Practical Guide to
Support Vector Classification”. A online guide for
using LIBSVM tools, 2009.
[6] Pallis.G,Vakali A.”Insight and Perspectives for Content
Delivery Networks”. Communication of ACM
(CACM) 2006; 49(1):101-6.
[7] Ten, W, Chang,CY,& Chen,MS.(2005).”Integrating
Web Caching and Web Pre fetching in Client Side

__________________________________________________________________________________________
Proxies". Parallel and Distributed systems,IEEE
Transactions on 16(5),444-455.
[8] R.Sudhakarapandian,”Modified ART1 Neural
Networks for cell formation using production
data”.Key Bridge Marriott,Washington DC,USA
August 23-26,2008.
[9] Akshay Shenoy,”Improving the Performance of a
Proxy Server using Web Log Mining”.San Jose State
University,4-1-2011.
[10] Ali W, Shamsuddin S M, Ismail A S, “Web Proxy
Cache Content Classification Based on Support Vector
Machine”, Journal of Artificial Intelligence 4 (2011)
100–109..
[11] Chen X, Zhang X. “Popularity-based PPM: an
effective Web Prefetching technique for high accuracy
and low storage.” In: Proceedings of the international
conference on parallel processing. Canada,
Vancouver; 2002.
[12] Chen Y, Qiu L, Chen W, Nguyen L, Katz R H.
“Efficient and Adaptive Web Replication using
Content Clustering.” IEEE J Selected Areas
Communication 2003; 21(6):979–94.
[13] Pallis G,Angelis L,Vakali A.”Validation and
Interpretation of Web Users’ Session
Clusters.”Information Processing and Management
2007; 43(5):1348-67.
[14] Web reference: http://www.wikipedia.com/svm..
[15] Ali W,Shamsuddin S M “Intelligent Client-Side Web
caching Scheme based on least recently used algorithm
and neuro-fuzzy system”.in:W.Yu,H.He,N.Zhang
(Eds.),Advances in Neural Networks-ISSN
2009,Springer,Berlin/Heidelberg,2009,pp.70-79..
[16] Yang Q,Zhang H.”Integrating Web Pre-fetching and
caching using prediction models”.World wide web
2001;4(4):299-321.
[17] Chen T,”Obtaining the optimal cache document
replacement policy for the clustering system of an EC
website”,European Journal of Operational Research
181(2007)828-841.
[18] cherkasova L,”Improving WWW Proxies Performance
with greedy-Dual-Size-Frequency

Enhancing proxy based web caching system using clustering based pre fetching with machine learning technique

More Related Content

Viewers also liked (20)

Similar to Enhancing proxy based web caching system using clustering based pre fetching with machine learning technique (20)

More from eSAT Publishing House (20)

Recently uploaded (20)

Enhancing proxy based web caching system using clustering based pre fetching with machine learning technique