基于项目聚类和评分的时间加权协同过滤算法

Time-weighted collaborative filtering algorithm based on item clustering and item score

针对传统协同过滤算法中面临稀疏项目评分矩阵计算耗时不准确、同等对待不同时间段用户的项目评分这些影响推荐精度的问题,提出了基于项目聚类和评分的时间加权协同过滤推荐算法(TCF)。该算法将项目评分与项目属性特征综合相似度高的聚到一个类别里,能有效解决数据稀疏性问题,降低生成最近邻居集合时间。引入时间加权函数赋予项目评分按时间递减的权重,根据加权后的评分寻找目标用户的最近邻居集合。实验从平均绝对误差、平均排序分和命中率三个指标来表明改进算法能有效提高推荐的准确性。

The calculation of traditional collaborative filtering algorithm is time-consuming and not accurate when faced large data of user project set, same treatment to user＇s item scores in different period also affect the accuracy. Therefore this paper presented the collaborative filtering algorithm based on item clustering and time-weighted item score （TCF）. Firstly ,this algorithm took the high comprehensive similarity with item score and characteristics into a category could effectively solve the problem of data sparsity and reduced the time to generate a set of the nearest neighbors. Secondly, it proposed time-weighted function gave the item time descending weights, and then used the weighted score to find the target user＇s nearest neighbors set. Experiment from three indicators：the mean absolute error, mean sorting point and hit rate all show that improved algorithm can effectively improve the recommendation accuracy.