Using Expert based Preference Elicitation for Collaborative Filtering Recommender systems
Traditional collaborative filtering assumes that user provides rating for any number of items given, and the system can infer the preferences. However, this approach proved to be inefficient and unrealistic. The problem of sparsity is amplified for cold start users which have an impact on the quality of recommendation. A new approach is proposed in this paper that identifies only the selected minimum list of informative items for which the user can provide ratings and thereby the system can generate personalized recommendations. In order to realize such a symbiosis, we need to consider the users who are recommended to items to rate, i.e. they have the potential and interest to rate the items. The selection of items considers the experts based opinion that helps the system in determining the users interest. By activating the users, we are addressing the problem of data sparsity. Our proposed system limits the number of input items for the user to rate and capture valuable output preferences. Results prove that our proposed strategy is better than the state of art strategy.