摘要
用户相似性和最近邻集合是协同过滤算法中最重要的两个步骤。传统的协同过滤算法依靠用户评分计算用户相似性并寻找K个邻居作为最近邻的方法为用户产生推荐,但是在数据稀疏的情况下,仅仅依靠用户评分使得推荐效果不准确。针对以上问题,文中提出一种改进的面向数据稀疏的协同过滤推荐算法。该方法引入用户属性相似性和用户兴趣度相似性,并结合传统的用户评分相似性计算用户间的相似度,通过多次实验调整三者的权重,并且采用动态选取邻居集合的方法确定用户的最近邻,从而为用户推荐最合适的项目,增强了方法实用性,以此来缓解用户数据稀疏性问题。实验结果表明,文中方法能够充分利用用户的各类数据信息,提高了预测评分的准确性及推荐质量。
User similarity and nearest neighbor set is two important steps in acollaborative filtering algorithm. The traditional Collaborative Filtering （CF） computes user similarity only relying on user rating and finds K neighbors as nearest neighbor to produce recommendation for users,but in the case of sparse data,only relying on user rating calculation makes the recommendation effect inaccurate. To solve the problems, an improved collaborative f＇dtering recommendation algorithm for data sparsity is proposed, which introduces the similarity of user attributes and user interest,combined with traditional user rating similarity to compute similarity between users. The weights of three is adjusted through several experiments,and the dynamic method is used to search the user＇s nearest neighbor to recommend suitable iterns for users,in order to alleviate user data sparsity problem. Experimental results show that this method can make full use of all kinds of users＇ data information .improving the accuracy of predicted ratings and ouality of recommendation.
出处
《计算机技术与发展》
2016年第3期63-66,共4页
Computer Technology and Development
基金
中央高校基本科研业务费专项资金资助项目(GK201002028
GK201101001)
陕西师范大学学习科学交叉学科培育计划资助项目
关键词
用户相似性
属性
兴趣
动态
数据稀疏性
user similarity
attribute
interest
dynamic
data sparsity
