An investigation on enhancement of fastness and scalability using semantic based clustering for top n recommendation system

Abstract

Search engines are more popular to explore the online services through Internet. In business aspect, studies of customer interest and past purchasing history have a great influence on improvement of producers. This is because of today s practice of online users preferring to get a suggestion from trusted parties before availing a service or a product which is called as recommendation. In dynamic environment, it suffers from data sparsity and scalability issues. As a contribution, the work proposes method to improve fastness and scalability of recommendation system for web service with QoS property. In recommendation, tracing similar neighbours is important process newlinefor which the searching space is reduced using clustering technique. Only the cluster to which active customer belongs are its neighbours instead of searching whole dataset. The first pre process work is dimensionality reduction which selects only important discriminate terms for web documents. Ontology based feature selection extracts semantically related terms .The sum of the link count and frequency decides the weight of the term. The documents are clustered using hybrid of k-means and seed based Cover Coefficient Clustering (C3M) methodology. Based on document-document semantic coverage value of C3M algorithm, the resultant semantic clusters form a community of highly bonded members. In the second work, the clusters result is further replenished by semantic labels as a profile of all members. It reveals the common background knowledge of the community to narrow the search space there by reducing searching time. Any ancestor node that subsumes a concept node is its relevancy node. The terms are looked up in Word net ontology for grasping common and root hypernyms. newline newline