融入知识图谱的电影推荐算法

Algorithm for Movie Recommendation Based on Knowledge Graph

传统的电影推荐系统可为用户推荐感兴趣的电影,但存在数据稀疏和冷启动问题。本文提出了ALS-G算法,将知识图谱作为辅助信息融入推荐系统中。首先利用图卷积神经网络GCN对电影属性知识图谱进行卷积获得电影的向量,将该电影向量与用户向量作为输入,然后利用ALS推荐算法预测该用户对该电影的喜爱程度。在MovieLens-1M数据集上的验证结果表明,本文的方法不仅能够缓解数据稀疏和冷启动问题,而且还提高了推荐系统的准确性。

Raditional movie recommendation systems can recommend movies of interest to users,but they have issues with sparse data and cold start.This paper proposes the ALS-G algorithm,which integrates the Knowledge graph as auxiliary information into the recommendation system.First,the graph Convolutional neural network GCN is used to convolve the film attribute Knowledge graph to obtain the film vector,and the film vector and user vector are used as inputs.Then,the ALS recommendation algorithm is used to predict the user's preference for the film.The validation results on the MovieLens-1M dataset show that our method not only alleviates data sparsity and cold start issues,but also improves the accuracy of the recommendation system.