摘要
提出了一种基于项目综合相似度的协同过滤算法。综合相似度是项目相似度和类别相似度进行加权,加权方式是从热能学中协同计算燃烧传热量的高温辐射换热综合发射率ε公式比拟得出,两者均是计算综合系数,在计算综合系数中可以通用。实验结果表明,在推荐不同的前N个项目的实验中,用新方法得到的准确率高于传统方法;在固定推荐数目改变最近邻的实验中,用新方法得到的准确率高于传统方法,因此可以得出结论:基于项目综合相似度的协同过滤算法可以提高计算准确性,提高推荐质量。
Abstract ： Proposing a new collaborative filtering algorithm based complex similarity of the item. The similarity is an item simi- larity and category similarity weighting. It derived the algorithm from the energy science emissivity e formula which came from the collaborative computing combustion heat of the energy science. Both of them wei＇e calculated complex coefficient which could be common in the calculation of the coefficient. Experimental results show that the accuracy obtains with the new method is higher than traditional methods in the experiments of recommending N items. A higher accuracy obtains by using a new method in the fix recommendation and change numbers of the nearest neighbors in the experiment. Therefore it can be conclu- ded that using the complex similarity based the item can improve the accuracy and improve the quality of recommendation.
出处
《计算机应用研究》
CSCD
北大核心
2014年第2期398-400,共3页
Application Research of Computers
关键词
协同过滤
项目相似度
类别相似度
综合相似度
发射率
collaborative filtering
item similarity
category similarity
complex similarity
emissivity
