Rating prediction of recommended item based on review deep learning and rating probability matrix factorization

作者：

Highlights：

• Combines the Deep Learning for review texts and the Probability Matrix Factorization method for rating data to predict the rating of the recommended items accurately.

• A deep learning framework of RNN with bi-directional GRU was designed to learn deep and nonlinear features of user preference and item characteristics from review documents.

• Experimental results validate that the proposed model performed better than the other state-of-the-art models.

• The proposed model has obvious cold start alleviation effects of the integrated review texts.

摘要

•Combines the Deep Learning for review texts and the Probability Matrix Factorization method for rating data to predict the rating of the recommended items accurately.•A deep learning framework of RNN with bi-directional GRU was designed to learn deep and nonlinear features of user preference and item characteristics from review documents.•Experimental results validate that the proposed model performed better than the other state-of-the-art models.•The proposed model has obvious cold start alleviation effects of the integrated review texts.

论文关键词：Precise recommendation,Ratings and reviews,Deep learning,Bi-directional GRU network,Self-attention mechanism,Probability matrix factorization

论文评审过程：Received 13 October 2021, Revised 7 May 2022, Accepted 22 May 2022, Available online 25 May 2022, Version of Record 8 August 2022.

论文官网地址：https://doi.org/10.1016/j.elerap.2022.101160