Semantic-based Web Mining Approach for Solving First Rate and Sparsity Problems in Recommendation Systems

Web recommender systems are used in different domain applications to predict customer preferences. They also used to assist the web users to satisfy their search in different products and services. Nowadays, a web system with effective and reliable recommendation techniques has become the target of the research in recent years. Number of methods has been proposed to the users for effective and efficient recommendations ranging from traditional approaches to the recent sophisticated web mining techniques; still some drawbacks are present in the current recommender systems. A frame work have been proposed to improve the scalability and sparsity of the recommendation techniques, starting from collaborative filtering approach to the first rater and cold start problems. The frame work is addressed to any online product recommendation system and can be used or extended to other products/domains. It uses different predictive model techniques for recommendations based on different situations. The input data for these techniques are the product attributes and the user attributes integrated with various data mining algorithms structured according to particular domain ontology. In this paper, a recommendation framework is proposed to overcome the major drawbacks of the current recommender systems in buying products.