Abstract: The traditional matrix factorization model can not effectively extract the features of users and items, while the deep learning model can extract the feature information well. At present, the mainstream recommendation algorithm based on deep learning only uses the output of neural network or the product of item features and user features to make recommendation prediction, which can not fully mine the relationship between users and items. Based on this, this paper proposes a recommendation algorithm based on the combination of text convolutional neural network and bias singular value decomposition (BiasSVD). Text convolutional neural network (TextCNN) is used to fully extract the feature information of users and items, and then singular value decomposition method is used to make recommendations, which can deeply understand the document context information and further improve the accuracy of recommendation. After extensive evaluation and analysis on two real datasets of MovieLens, the recommendation accuracy of this algorithm is obviously better than that of ConvMF algorithm and mainstream deep learning recommendation algorithm.

Key words: matrix decomposition, singular value decomposition, deep learning, text convolutional neural network