Collaborative Filtering Improving Information and Knowledge - Methods for Spreading the Field of Collaborative Filtering

摘要

A collaborative filtering is useful way for the information and knowledge sharing. The collaborative filtering supports the information and knowledge sharing based on user's similarity, which are discovered by the data such as user's history. We are studying that the introducing of collaborative filtering to improve the information and knowledge sharing more efficiently than ever. To spread the applicable field, the existing collaborative filtering must solve following problems: (1) "the Cold Start Problem", (2) improvement in the speed of processing and (3) improvement in the prediction accuracy. Objective of this research is to provide the methods that can solve these three problems. In this paper we propose the presumption method of a deficit value to the Cold Start Problem, the extraction method of a useful data to the improvement in the speed of processing and the discovery method of a similarity among user to the improvement in the prediction accuracy. Additionally we conducted the comparison experiments using these proposed methods and confirmed the efficiency.