面向电影推荐的时间加权协同过滤算法的研究

发布时间：2018-09-19 10:44

【摘要】：针对协同过滤算法的信息过期问题,提出一种改进的时间加权协同过滤算法(NTWCF)。考虑信息随时间推移导致信息影响力变化的因素,在信息半衰期的启发下,引入信息保持期的概念,通过在最近邻查找阶段和预测评分阶段采用一种新颖的时间加权函数为项目上的评分赋予不同的时间权重。电影数据集上的实验结果表明,它在一定程度上大幅度提高了预测推荐的准确性。接着,针对算法的实时性问题,利用时间加权的项目聚类优化NTWCF算法,提出综合时间权重和项目聚类的协同过滤算法(TWICCF),对评分信息时间加权后再对项目K-means聚类,在为目标项目查找最近邻时只在若干聚类构成的项目集中进行。相比NTWCF算法,TWICCF算法在推荐准确度和实时性上均取得了显著的提升。
[Abstract]:Aiming at the information expiration problem of collaborative filtering algorithm, an improved time-weighted collaborative filtering algorithm (NTWCF). Is proposed. Considering the factors that lead to the change of information influence over time, and inspired by the half-life of information, the concept of information retention period is introduced. By using a novel time weighting function in the nearest neighbor lookup stage and the prediction scoring stage, different time weights are assigned to the item score. The experimental results on the film data set show that it improves the accuracy of the prediction recommendation to a certain extent. Then, aiming at the real-time problem of the algorithm, using the time-weighted item clustering optimization NTWCF algorithm, a collaborative filtering algorithm, (TWICCF), which synthesizes the time weight and item clustering, is proposed to cluster the item K-means after weighted time of scoring information. Finding nearest neighbors for the target project is only done in a cluster of items. Compared with NTWCF algorithm, TWICCF algorithm has achieved significant improvement in recommendation accuracy and real time.
【作者单位】：大连东软信息学院软件工程系;大连理工大学软件学院;
【基金】：面上基金:在线背包问题的算法和分析(11101065)资助
【分类号】：TP391.3

【参考文献】