Efficient Techniques to Improve Clustering Accuracy on Web by Using Hybrid Approach

Author Name(s): Monika Sharma, M A Rizvi
Author Email: moni.sharma357@gmail.com

Abstract

Clustering in web mining has become a challenging task in the present scenario, which draws the attention of many researchers. Cluster analysis is mainly used to determine the clusters and algorithms for identifying the best cluster. Cluster analysis includes a number of unsupervised order techniques that are planned to discover a framework in datasets to form clusters where comparable information objects are gathered together into a well-organized groups. World Wide Web is a rich information source, along with a large number of the web users extracting information using web are increasing. The weblog records are increasing rapidly in a disorganized manner, it contains a lot of information and the major problems of internet users are to extract meaningful information and to discard irrelevant data. In this paper, various clustering methods are examined to recognize its significance with regards to the large dataset and observationally these have been tried on Wine Quality dataset taken from UCI repository to feature their qualities. In this paper, an attempt is made to develop a new hybrid technique is introduced to provide better clustering accuracy.

Introduction

Recently, clustering in web mining gained a lot of attention and becomes a most popular technique for performing its main task to extract the interesting patterns from the internet. A cluster is characterize as a group of items that have a more advanced level of similarity one and all are correlated with objects that are not in the similar set. Still, there is an uncertainty respecting reasonable closeness metric for clustering. Various measures have been recommended to evaluating the closeness, i.e. Euclidean distance, density in data space and so on. The internet presents a lot of learning and data for users to access various types of information as per the user’s choice. But, the web is a collection of relevant as well as irrelevant information. Retrieving the significant data from the web turns out to be more complex to the users. To get rid of this problem, clustering technique is introduced.

Conclusion

In this research paper, we target to produce a good quality of clusters by evaluating the comparison between various algorithms. As result shows that hybrid hierarchical clustering algorithm performs better as compared to other algorithm i.e. agglomerative hierarchical clustering. Results prove that accuracy is approximately 79% achieved by average method in hybrid hierarchical clustering algorithm. It is more efficient for clustering the web pages. The proposed techniques can also be evaluated on different datasets.

874 total views, no views today

Download PDF File

About the author: admin