Web Document Clustering for Finding Expertise in Research Area

Researchers often need to find expertise in their chosen area of research. Finding expertise is very useful as relevant research papers can be studied and the experts could be identified. Therefore finding expertise in the chosen area of research has always attracted interest among academic community. These days research institutions and individual researchers make their publications and research findings available on web. With the exclusive growth of World Wide Web search engine users are overwhelmed by the huge volume of results returned in response to a simple query, which is far too large to get the desired knowledge. Therefore one of the methods of finding the expertise is by way of efficiently and accurately clustering the web documents, which enhances the integrity of web search engine. Data mining techniques matured making it possible to automate the web document clustering. In this paper, we present mutually exclusive Maximal Frequent Item set discovery based K- Means clustering approach. It has been implemented in JAVA. The common text processing approach is to convert the downloaded web documents into vectors. It is being done by extracting document features and it generates the document-feature data set. For a set of documents, the feature set is composed of all terms appearing in any one of the documents. We call this a document-feature data set. If document m contains feature n, then the corresponding value, in row n and column m of the table, is set to one. Otherwise, it is zero. Then, Apriori algorithm is applied to these document feature data set. The mutually exclusive frequent sets generated by Apriori algorithm are taken as initial points of K-Means algorithm. The output of the K- Means clustering algorithm will be the sets of highly related documents appearing together with same features. This approach enables the clustering of the web documents. It enables researchers to find the documents related to their desired area clustered and displayed together during the web search. It will significantly help them in terms of saving the time and getting all the relevant papers together in a cluster..