推荐系统的信息核挖掘及其应用研究

项目名称： 推荐系统的信息核挖掘及其应用研究

项目编号： No.61502078

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 计算机科学学科

项目作者： 曾伟

作者单位： 电子科技大学

项目金额： 21万元

中文摘要： 推荐系统已经在理论和应用中取得很大进展，但其所需处理的数据往往规模巨大，从而导致推荐算法的效率不能满足应用需求。本项目拟基于复杂网络理论，研究基于二部图网络表示的推荐系统，挖掘推荐系统的信息核，在满足算法准确性要求的前提下，大大提升推荐算法的效率。首先，从推荐网络拓扑结构的统计特征入手，挖掘推荐网络的信息核，设计信息核提取算法，确保信息核在满足推荐算法功能需求的前提下极大化压缩数据量；其次，通过信息核的演化分析，建立模型刻画其网络结构的动态演化机制，力求通过信息核网络的演化特征揭示原始推荐网络的演化特征，进而实现推荐网络建模；最后，设计基于信息核的静态和动态算法，利用推荐网络信息核保功能而规模极小的优势，设计高效的推荐算法。本项目在保证推荐功能的前提下对推荐网络进行结构压缩，为推荐算法处理大规模数据集提供新思路，其结果不仅有利于在应用实践中取得效益，而且丰富了复杂网络理论研究。

中文关键词： 推荐系统；协同过滤；复杂网络；网络演化

英文摘要： Recommender systems have made great progress in the theory and application. However, a recommender system usually has a large number of users and items, which make a recommendation algorithm dissatisfy requirements of real applications. This project aims to improve the efficiency of recommender systems by uncovering the information core in recommendation networks based on the theory of complex network. Firstly, we will investigate the property of recommendation networks to uncover the information core, and then propose recommendation algorithms which only use the information core data but generate satisfactory recommendations. Secondly, we will study the evolution characteristics of the information core network and propose models. Our goal is to reveal the evolution characteristics of the origin recommendation network by the evolution of the information core network. Finally, with the information core, we will design static and dynamic algorithms whose efficiency will benefit from the advantage of the information core. In this project, we compress the recommendation network on condition that the performance of the recommender system is preserved. Our project provides a novel way to solve the big data problem in recommender systems. Our results have great significance in real applications as well as the research of the complex network.

英文关键词： recommender system ;collaborative filtering;complex network ;network evolution