基于改进协同过滤的视频推荐算法研究

Research on Video Recommendation Algorithm Based on Improved Collaborative Filtering

针对视频推荐领域因用户评分数据极度稀疏以及推荐不准确而导致的用户体验度低的问题,论文提出一种基于网络用户兴趣和视频标签的算法。首先,将用户具体行为数据纳入考量,为行为赋值,用用户兴趣矩阵取代用户项目评分矩阵。接着对视频标签进行分析,引入权重因子,加权计算视频相似度,获取相似视频候选集。最后对目标用户进行TopN推荐。经过实验,论文的改进算法准确率在MovieLens数据集上提高了15%。

Aiming at the problem of low user experience caused by extremely sparse user rating data and inaccurate recommendation in the field of video recommendation,this paper proposes an algorithm based on user interest and video tags.Firstly,the user's specific behavior data is taken into account,and the user's item rating matrix is replaced by the user's interest matrix.Then,the video tags are analyzed,and the weight factor is introduced to calculate the video similarity to obtain the similar video candidate set.Finally,TopN is recommended to the target users.Through experiments,the accuracy of the improved algorithm is improved by 15%on MovieLens dataset.