Recommendation Model Based on Asymmetric Neural Matrix Factorization
Hangyu Yan, Yan Tang, Lei Yan · 2019
In recent years, deep learning has been combined with traditional recommendation system methods to achieve better recommendations. The existing deep matrix factorization recommendation model combines implicit feedback with explicit score to recommend. Although this combination improves the recommendation effect, it fails to make full use of explicit score and implicit feedback. On this basis, this paper proposes an asymmetric neural matrix factorization recommendation model. By constructing a score matrix that can make full use of explicit score and implicit feedback, and using asymmetric neural network to learn the matrix. The new model and the original matrix factorization model were compared on two public datasets, evaluated by HR and NDCG, indicating that the asymmetric neural matrix factorization model has been improved in recommendation.