Enhancing Social Recommenders with Implicit Preferences and Fuzzy Confidence Functions.

In this paper we explore the use of fuzzy confidence functions for enhancing the performance of social recommenders with implicit preferences, focusing on K-Nearest Neighbors (KNN) collaborative filtering algorithms. Firstly, we measure the effects of including social relations for enhancing the performance of the algorithms with either explicit or implicit preferences, expecting to verify better results when social attributes are considered in the relevant neighborhood estimation. Secondly, it is proposed to enhance the social recomenders with implicit preferences by fuzzy confidence functions. An application is developed to measure the effects of including social relations and to illustrate our proposal on the fuzzy modeling of implicit preferences, recommending courses based on the students socio-demographic and academic information. As a result, the best recommendations are accomplished with socially-enhanced algorithms that make use of implicit preferences and fuzzy confidence functions, obtaining a FCP in test of 0.68.