Collaborative Filtering Recommendation Algorithm of Behavior Route Based on Knowledge Graph

For personalized recommendation, the commonly used recommendation algorithms include content recommendation, item collaborative filtering (Item CF) and user collaborative filtering (User CF), but most of these algorithms and their improved algorithms tend to focus on the explicit feedback of users (tags, ratings). Etc.) or scoring data, lack of utilization of multi-dimensional user behavior and behavior sequence, resulting in insufficient recommendation accuracy and cold start. In order to improve recommendation accuracy, this paper proposes a behavior path collaborative filtering recommendation algorithm based on knowledge graph (BR-CF). First, create a behavior graph and behavior route based on user behavior data, considering the sequence of behaviors, and then use vectorization technology (Keras Tokenizer) to vectorize the text-type path, and finally calculate more The similarity between the dimensional behavior path vectors is to perform collaborative filtering recommendations for each dimension. On this