基于复杂结构数据聚类的推荐系统

A recommender system based on clustering of complex structured data

针对目前推荐系统存在的不能处理结构复杂、语义丰富领域的推荐问题以及对项目空间和用户空间本质特征理解的狭窄性和简单性、稀疏性问题、可扩展性问题，研究了基于复杂结构数据聚类的推荐方法，提出了一个新颖、有效、具有高可扩展性的基于复杂结构数据聚类的混合型推荐系统HRSCCSD。该系统能同时融合用户语义、项目语义和项目协同多方面信息，极大地拓展了当前推荐系统的应用深度和广度。实验表明，所提出的推荐技术在覆盖性、准确性以及可扩展性方面均优于当前主流的推荐技术。

The HRSCCSD, a novel, elegant and scalable hybrid recommender system based on clustering of complex structured data is proposed. The system can resolve the problems of the current recommender systems such as unfeasibility in complex structured and rich semantic fields, unilateralism and simplicity for employing item space and user space, sparsity, and scalability, and take account of semantic information about users, semantic information about items, and collaborative information about items. As a result, the application in recommendation can be greatly expanded. The experimental results indicate that the proposed recommendation method is superior to the current mainstream techniques in the respects of covergae, accuracy and expansibility.