【摘要】 协同过滤算法作为推荐系统中一种常用算法,在实际应用中还存在一些问题,如传统协同过滤算法里对称相似度计算方法会导致用户相似度测量值存在误差.针对这个问题,提出一种非对称的加权相似度协同过滤方法.通过计算用户共同评分项目所占比例来确定用户相似度非对称加权因子,以表现用户之间相互影响的差异性;通过加权因子和传统相似度度量方法确定用户相似矩阵,使用矩阵分解梯度下降法来拟合没有共同评分项目的用户之间的相似度数据.最后,通过在Movie Lens和Douban数据集进行实验验证和对比,以均方根误差和平均绝对误差作为评判标准,实验结果表明本文所提方法的推荐准确度更高.更多还原

【Abstract】 As a common recommendation system algorithm,there are still some problems in the Collaborative Filtering algorithm in practical applications,such as the user similarity measurement errors introduced by symmetrical similarity calculation method in the traditional collaborative filtering algorithm.For this problem,we propose an asymmetric weighted similarity collaborative filtering methods.The method determines the similarity asymmetric weighting factor by calculating the proportion of user common rating project,so as to showthe interaction differences between the user,and determines the user similarity matrix by using matrix decomposition gradient descent algorithm to fit the similarity score data between the user which has no common rating project.Finally,we have a verification and comparison on the Movie Lens and Douban data sets in the experiments.By using the root mean square error and the mean absolute error as the criterion,the experimental results showthat proposed method of this article has a higher recommendation accuracy.更多还原