摘要
为了提高用户相似度计算的精度和知识推送的准确度,提出了一种基于热门物品惩罚和用户兴趣变化的协同过滤推送算法。该算法首先对知识项进行聚类;其次在每一类中引入用户兴趣度函数来对类内未评分知识项进行评分值预测;然后在每个类的用户相似度计算中引入热门物品权重系数,用以惩罚热门物品对用户相似度的影响;最后在推送当中引入用户兴趣随时间变化的权重系数。实验还采用MovieLens数据集进行了测试,结果表明,改进后的算法比传统的协同过滤算法在推送准确度上有明显提高。
In order to improve the accuracy of user similarity calculation and the accuracy of the push of knowledge, this paper puts forward a collaborative filtering push algorithm based on hot item punishment and user interest change. Firstly, the algorithm is utilized to cluster items into several classes. Then, in each of class the user＇s interest degree function is introduced to compute the predictive values in each class; and then in each class user similarity calculation hot items weight coefficient is introduced to punish the similarity influence of hot items on user similarity. At last, in the final recommendation algorithm time weight coefficient of user interest change is introduced. The experiment also uses MovieLens data set to test the algorithm, and the results show that the improved algorithm is better than the traditional collaborative filtering in the push on accuracy.
出处
《系统工程》
CSSCI
CSCD
北大核心
2014年第1期118-123,共6页
Systems Engineering
基金
国家自然科学基金资助项目(71172169)
关键词
协同过滤
热门物品惩罚
知识项聚类
用户兴趣变化
Collaborative Filtering
Hot Item Punishment
Knowledge Item Clustering
User Interest Change
