Research on items Recommendation Algorithm Based on Knowledge Graph

Traditional recommendation systems mostly use collaborative filtering algorithm, which has problems with cold start and data sparseness. A common idea to solve these problems is to introduce some auxiliary information as input in the recommendation algorithm. The knowledge graph contains rich semantic information, which can provide potential assistance for the recommendation system. The research on the existing recommendation methods based on knowledge graphs found that these methods lacked the consideration of entity attributes information. Therefore, this paper considers attribute factors and proposes the interest modeling method the entity attributes-based in the knowledge graph, and fusions with traditional collaborative filtering algorithm to improve the recommended effect. The results show that the proposed recommended algorithm has better property than other commonly used benchmark algorithms.