Abstract: Most of the traditional recommender systems emphasize on accuracy, while ignoring the diversity, and the data set lacks of accessorial information. A recommendation algorithm incorporating user preferences and network structure is proposed. Firstly, the user’s historical feedback data are used to analyze the diversity preferences of the user. Pre-recommended list incorporating the user preferences and BGPR（Bipartite Graph Projection and Ranking）can be obtain. Then collaborative tags contain abundant information about personalized preferences and item contents, and are potential to mine the user’s favorite tags to obtain indirect related recommended list. Finally, the model is employed to generate the final diversified recommendation list. The experimental results show that the proposed method can effectively improve the diversity of recommended list under the premise of ensuring the accuracy rate.

Key words: recommender system, user preferences, diversity, network structure

摘要： 针对传统推荐系统追求推荐列表的准确率而忽略推荐的多样性以及数据集信息缺失等问题，提出了融合偏好度与网络结构的推荐算法。通过用户历史反馈数据分析用户偏好度，将偏好度与二部图随机游走推荐算法融合，初步得出项目推荐列表；利用用户-标签二部图，挖掘用户不跟随大众的喜好标签，得到推荐项目列表；根据模型融合得到最终的推荐结果。实验表明，新算法在保持较好精确率和召回率的情况下，有效提高了推荐的多样性。

关键词: 推荐系统, 偏好度, 多样性, 网络结构