Collaborative filtering hybrid recommendation algorithm incorporating knowledge graph

Abstract

Nowadays, in the context of the booming digital economy and video becoming the main carrier of data explosion, enhancing the precision of recommendation algorithms in the video industry has emerged as a prominent area of investigation. By using the TransR model to construct the movie knowledge graph into the relationship space to obtain the movie entities and their relationships, the multiple relationships between movies are better reflected, so as to calculate the semantic similarity between movies, and then the collaborative filtering algorithm based on Pearson coefficient calculates the similarity of user behavior, and the two similarities are linearly fused to finally generate the final recommendation list for Top-N recommendation. Comparative experimental results show that the algorithm has improved in the main indexes, such as recall, accuracy, and mean absolute error (MAE).