摘要
针对现有推荐算法忽视了推荐结果总体多样性的问题,在传统的协同过滤算法基础上,提出了新的算法。分析了基于项目的协同过滤算法的原理及多样性缺陷,有针对性地对其改进。该方法降低了活跃用户、热门商品对计算商品相似度的贡献,并利用贝叶斯理论分析用户对商品特征属性的喜好度。在计算相似度时,考虑用户对商品特征的喜好度,在此基础上计算目标商品的最近邻居。实验结果表明该算法可以有效提高推荐系统的多样性。
Aiming at the problem that the existing recommendation algorithms ignore the overall diversity, a new algorithm is presented based the traditional collaborative filtering algorithm. This paper analyzes the principles of item - based collaborative filtering algorithm and the defects of diversity, targeted for its improvement. The new algorithm reduces the contribution of active users and hot products on calculating item similarity, and analyzes the the user＇s preferences of the item characteristics by using Bayesian theory. It calculates item similarity degree considering the user＇s preferences, and then acquires nearest neighbors of the target item. Experimental results show that the new algorithm can effectively improve the diversity of the recommended system.
出处
《无线通信技术》
2013年第3期5-9,共5页
Wireless Communication Technology
基金
江苏省普通高校研究生科研创新计划(1221170028)
江苏省高校自然科学基金资助项目(10KJB520004)
关键词
协同过滤
多样性
贝叶斯理论
商品相似度
collaborative filtering
diversity
bayesian theory
item similarity
