Abstract: Aiming at the problem of recommender system facing the data sparseness, cold-start, user features dynamic change and different users focus on different features, this paper puts forward a recommendation algorithm that combines user characteristics classification and the dynamic time. To compensate for the lack of computational similarity, it considers the dynamic change of user characteristics and classifies users based on feature, and solves the problem of cold start combining the user’s feature and the user’s score. The results show that the proposed algorithm can effectively improve the quality of recommendation.

Key words:  recommender system, cold-start, user characteristics, dynamic time, similarity calculation

摘要： 针对推荐系统中数据稀疏、冷启动和用户特征动态变化及不同用户对同一特征依赖程度不同等问题的影响，提出了结合用户特征分类和动态时间的协同过滤推荐。考虑用户特征动态变化的同时将用户依据特征分类以弥补计算相似度的不足，并将用户特征和用户评分相结合解决冷启动问题。结果表明该算法能有效提高推荐质量。

关键词: 推荐系统, 冷启动, 用户特征, 动态时间, 相似性计算