基于动态时序的特征复合协同过滤算法

Feature Composite Collaborative Filtering Algorithm Based on Dynamic Timing

协同过滤算法是推荐系统中研究较为广泛和深入的算法,为解决传统协同过滤算法无法处理时间动态变化的问题,提出一种新的改进算法:SpecialTSVD++算法。在传统SVD++算法中分别融入用户评分的时间信息、用户和物品的时间偏置,并且加入用户特征信息,增强数据与时间的关联度,体现数据的动态变化,并且结合用户属性产生个性化推荐结果。Movielens-10m数据集上的实验结果表明,SpecialTSVD++算法通过对时间动态变化带来的推荐影响进行优化处理,使推荐结果更加贴近用户当前需求,能显著提升推荐系统准确率。

Collaborative filtering algorithm is a widely studied and in-depth algorithm in the recommendation system,but the traditional collaborative filtering algorithm can’t meet the dynamic changes in the future time,So the algorithm of this article integrates the time information of the user's rating,the user's characteristic information,and the time bias of the user and the item into the traditional SVD++algorithm.And create a new and improved algorithm:SpecialTSVD++algorithm,which enhances the correlation between data and time,reflects the dynamic changes of the data,and generates personalized recommendation results in combination with user attributes.The experimental results on the Movielens-10m dataset show that the SpecialTSVD++algorithm of this paper optimizes the impact of the dynamic changes in time,and the recommendation results are closer to the user's current needs,which significantly improves the accuracy of the recommendation system.