A Personalized Recommendation System based on Knowledge Graph Embedding and Neural Network

The application of Neural Network to recommendation task has gradually drawn attention over the last few years, and a recommendation algorithm combining neural network with collaborative filtering has emerged. Meanwhile, knowledge Graph and Graph Embedding have also developed considerably. In this paper, a new algorithm level solution is presented to realize personalized recommendation that is based on Knowledge Graph Embedding and Neural Network. Knowledge Graph Embedding is used to embed each entity into a low-dimensional vector. The learned vectors are as the input of the neural network to predict the score of an item. Through a series of systematic tests involving the MovieLens-1M dataset, we demonstrate that it can effectively improve the accuracy of rating prediction comparing with the original neural collaborative filtering algorithm.