Top-N recommendation algorithm for distance decomposition combined with biased deep learning

In view of the fact that traditional matrix factorization algorithms are mostly shallow linear models, it is difficult to learn the hidden feature vectors of users and items at a deep level, and the problem of overfitting is prone to occur in the case of sparse data. In this paper, a fusion bias depth is proposed. The learned matrix factorization algorithm can not only solve the problem of data sparseness, but also learn the distance feature vector with stronger characterization ability. First, the user-item interaction matrix is ​​constructed through the explicit and implicit data of the user and the item, and the interaction matrix is ​​converted into the corresponding distance matrix; secondly, the distance matrix is ​​input into the depth of the bias layer by row and column respectively. The neural network learns to obtain the distance feature vector of the user and the item with nonlinear characteristics; finally, the distance between the user and the item is calculated according to the distance feature