Extreme Residual Connected Convolution-Based Collaborative Filtering for Document Context-Aware Rating Prediction

• Published in: IEEE access 2020, Vol.8, p.53604-53613
• Main Authors: Zhang, Bangzuo, Zhu, Min, Yu, Mengying, Pu, Dongbing, Feng, Guozhong
• Format: Article
• Language: English
• Subjects: Artificial neural networks; Attenuation; Back propagation networks; Collaboration; collaborative filtering; Convolution; Deep learning; Feature extraction; Filtration; Kernel; Neural networks; recommender system; Recommender systems; Residual connection; Sparse matrices
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2020.2981088

Deep learning methods can improve the performance of recommender systems over traditional methods, especially when text information is available. To extract information hidden in the text description of an item and fuse it with rating information, it has been proposed that the network be stacked with more layers. However, as the network thus deepens, the problem of the attenuation of the preamble signal can cause the convolutional neural network to break down. The gradient gradually disappears during the back-propagation process, resulting in an inability to adjust the weights. In this paper, a method called eXtreme Residual connected Convolution Collaborative Filtering (xRConvCF) is proposed to predict the rating for each item based on the textual information. It creates data branch lines to form a residual module called the eXtreme residual (xRes) connection to mitigate the problem of the vanishing gradient and enhance feature reuse. The results of experiments on empirically obtained datasets show that the proposed deep learning model significantly outperforms state-of-the-art methods of recommendation.