Research on a Collaborative Filtering Method Introduced into Knowledge Graph

In our society, recommendation technology is more popular in Commercial, AI, and other related fields with the further development of the Internet. Information retrieval and filtering research is the initial origin of recommendation systems. The most used filtering type by RS at the beginning is collaborative, content-based, and demographic. Collaborative filtering (CF) is the most popular recommendation technology, it usually filters similarly interesting users' thoughts and choices to find aimed users' most likely future goods. However, the technical conditions and situation changed with the progressed technology and changed demands; the CF technology cannot be suitable for the latest conditions and situations. We introduce and apply knowledge graphs (KG) in CF technology to solve issues. The Knowledge Graph is a kind of technology that can reflect the structure relationship between different entities. Knowledge Graph attracts more attention from professional areas as a form of structural human knowledge, which constantly argues the user relationship's effects on recommendation results. This article finds three main measuring factors from Knowledge Graph and composed correlation vectors based on these factors: Relationship Strength Vector, Arrow Direction Vector, and Entity-relationship similarity vectors. To improve the CF recommendation systems, the Authors first use the latest three vectors to infer entity similarity, then compose with the traditional collaborative filtering technology to get a new method. Finally, there are some experiments to prove the effectiveness of this method. This method can further improve and change the recommendation order, letting the recommendation result has higher similarity with the actual users.