基于社交网络的单类协同过滤算法

One-class collaborative filtering algorithm based on social network

单类协同过滤算法的研究是当前协同过滤算法研究的重要课题,其面临的主要问题是:仅仅正例数据能够被观察到,类高度不平衡,大量的数据点丢失.把社会化正则项引入到传统的单类协同过滤算法,提出一种新的基于社交网络的单类协同过滤算法来解决这些问题.在真实的包含社交网络的数据集上实验验证,该算法在各个评价指标下性能均优于几个经典的单类协同过滤算法.

One-class collaborative filtering（OCCF） is an important task that naturally emerges in recommend dation system setting. Its typical problems included： only positive examples could be observed, classes were highly imbalanced, and vast majority of data points were missing. These problems influenced the performance of OCCF. With the advent of online social networks, we proposed a new OCCF algorithm based on social network to solve these problems by importing the social regularization term to classical OCCF. We conducted our experiment on a large real-world dataset with social information. The experiment results illustrated that our approach achieved a better performance than several traditional OCCF methods.