A particle filter approach for reputation information systems - Performance evaluations by a multi-agent simulation

This paper proposes a new method to evaluate content values from reputation information by using a particle filter. In reputation information sites such as Tabelog and Amazon.com, a user can post reviews and vote the evaluation of contents such as books, camera and restaurant. Reputation information sites estimate content values by statistically processing the evaluations which many users have voted, and provide an information for making a decision when many users to select contents. However, it is not always true that all users can vote the content values always correctly. So we need to correct the content value voted by users. In this paper, we investigate the efficiency of the proposed reputation information system through simulation experimentations.