Due to unprecedented growth of information on the Web and lack of structure in many Web sites, it became real challenge to the Web users to find relevant information. To solve this problem, Personalization is considered as a popular solution to customize the World Wide Web environment toward the user’s preferences. Recent study shows that Web Usage Mining techniques play an important role in designing Web page recommendation systems. However the present conventional content-based recommender systems, built using the Web Usage Mining process, do not take Semantic Knowledge into pattern discovery and recommendation process. Recent studies show that integrating domain knowledge in the form of ontology into Web Usage Mining process can enhance the quality of the discovered usage patterns. Our work aims to incorporate semantics knowledge in all the phases of Web Usage mining process. CloSpan, a state-of-the-art algorithm for Sequential Pattern mining is applied over the Semantic space to generate frequent Sequential Patterns. The generated semantically enriched patterns are fed to Web page Recommendation model in offline phase. Experimental results shown are promising and showed a significant improvement on the quality of the recommendations.
With the explosive growth of information on World Wide Web, it has become a real challenge for Web users to find and access relevant information. One possible approach to address this challenge is personalizing the user’s Web experience [1]. Web Personalization [2] is the process of customizing a Web site to meet the needs of each specific user or set of users, by taking advantage of the knowledge acquired through the analysis of the user’s navigational behavior. Web Recommender system is a specialized personalization system. Understanding the information needs of users has become a crucial task for Web site owners in improving, the services offered by them and to retain the customers. A key requirement in developing successful personalized web applications is to build user models that can accurately represent user’s interests and preferences. Further the user models should have the ability to encode the information in machine-understandable format for addressing semantic heterogeneity issues. There are various ways to model user needs. Web Usage Mining techniques are considered to be popular in modeling user’s web browsing behavior and thus patterns discovered from mining process are used in various applications such as building recommendation systems. Web Usage Mining is a process of extracting interesting and frequent navigational patterns by applying Data mining techniques on Web log files. However conventional Web usage based recommender systems are limited in their ability to use the domain knowledge of the Web application and their focus is only on Web usage data. As a consequence, the quality of the discovered patterns is low. These patterns do not provide explicit insight into the user’s underlying interests and preferences, thus limiting the effectiveness of recommendation system in interpreting and justifying the recommendations [3]. Recent studies [4] claim that Ontology, which represents the domain knowledge of the application in Semantic form, if integrated with Web Usage Mining can enhance the quality of generated usage patterns and help in building effective Recommendation systems. The combination of Web Usage Mining and Semantic Web has created a new and fast emerging research theme – Semantic Web Usage Mining [5]. The key contributions of our work are summarized as follows: 1) Building a model to formalize the user’s Web browsing activities in Semantic form. 2) Feeding domain Ontology into Web Usage Mining Process for extracting Sequential navigational patterns. Such integration allows more pruning of the search space in Sequential pattern mining of the web log file. 3) A state-of-the-art algorithm is used in the Sequential Pattern mining process to generate frequent Sequential Patterns and subsequently sequential association rules are generated. 4) Generated Sequential rules have antecedent and consequent as sequence of ontological instances instead of mere page views. The rest of the paper is organized as follows. In section II we review recent advances in the area of Semantic Web Usage Mining. In section III proposed model and architecture are discussed. Experimental set up and performance evaluation of the proposed model is presented in section IV. Finally section V provides the concluding remarks and sheds light on possible future enhancements.
Recent Comments