关键词: 好友推荐, 属性分类, 对象相似性

Abstract: Social network recommends friends according to the information of the users. This problem brings technical challenges. Current technologies cannot solve this problem effectively. Facing the challenges, this paper proposes a recommendation algorithm based on the classification of properties. It analyzes the contributions of different types of properties to users’ behavior by machine learning and classifies the types of results. Based on the classifications, it proves the algorithm can find more relevant friends or the merchandise which are more likely to arouse users’ interest so as to acquire the classification as well as the degree of closeness between users and their friends more correctly, rapidly and comprehensively. This paper validates the effectiveness and the efficiency of proposed algorithms with extensive experiments.

Key words: friend recommendation, property classification, object similarity