结合知识图谱进行信息强化的协同过滤算法

针对传统协同过滤算法存在使用信息单一、基础评分数据过于稀疏导致推荐效果不佳等问题,该文提出一种结合知识图谱进行信息强化的协同过滤(KGRI-CF)算法.该算法利用电影的特征数据构建1张关于电影的知识图谱,对用户-评分矩阵进行有条件的填充,有效改善了传统协同过滤算法的数据稀疏性问题.通过对评分数据进行统计与挖掘获取用户的偏好信息,构建了关于用户偏好的知识图谱.利用实体向量化算法将知识图谱中的实体以及关系向量化后计算出用户信息相似度,将其与基于用户的传统协同过滤算法得到的用户评分相似度以一定比例进行融合,从而得到最终的用户相似度,并以此为基础进行评分预测并得到推荐列表.实验结果表明:与传统协同过滤算法相比,该算法能有效地改善数据稀疏性问题,预测结果的精准率和召回率均有显著提升,同时具有较好的可解释性.

The Collaborative Filtering Algorithm for Information Enhancement Combined with Knowledge Graph

The collaborative filtering algorithm that combines knowledge graphs for information enhancement(KGRI-CF)is proposed,aiming at the problems of single use information and too sparse basic scoring data leading to poor recommendation effect in traditional collaborative filtering algorithms.The algorithm uses the feature data of the movie to construct a knowledge map about the movie,and conditionally fills the user-rating matrix,which effectively improves the data sparsity problem of the traditional collaborative filtering algorithm.Preference information is used to build a knowledge graph about user preferences.The entity vectorization algorithm is used to vectorize the entities and relationships in the knowledge graph to calculate the similarity of user information,which is fused with the similarity of user ratings obtained by the traditional user-based collaborative filtering algorithm in a certain proportion to obtain the final user similarity.On this basis the score prediction is performed and the recommendation list is obtained.The experimental results show that,compared with the traditional collaborative filtering algorithm,the algorithm can effectively improve the data sparsity problem,the accuracy and recall rate of the prediction results are significantly improved,and it has better interpretability.