关键词: 推荐系统, 相似性, 正态分布, 修正函数, 邻居用户集合

Abstract:

A calculation method of correlation similarity with normal distribution function as a correction function is proposed based on the traditional correlation for better accuracy of seeking the nearest neighbor set in the collaborative filtering recommendation system. The normal distribution function is employed to correct the negative effect because of the number of items between users on similarity calculation, which well reflects the degree of similarity of users. The experimental results show that the proposed method offer better accuracy of searching for the nearest neighbor set than the traditional method of correlation similarity under the same conditions.

Key words: recommender system, similarity, normal distribution, correction function, neighbor user set