Hybrid Collaborative Filtering with Semi-Stacked Denoising Autoencoders for Recommendation
Hairui Zou, Chaoxian Chen, Changjian Zhao, Bo Yang, Zhongfeng Kang · 2019
Recommender system is one of the solutions to deal with information overload problem, thus has been intensively studied. In recent research, side information has been commonly used besides the rating matrix, so as to mitigate the sparsity problem and to improve the recommendation accuracy. To better making use of side information, deep learning based recommendation methods have been proposed, among which autoencoder-based models have become quite popular. However, most existing autoencoder-based models require the each input corresponds to one output, which may bring in information loss and high cost to extend autoencoders, thus affects the recommendation accuracy. To address this important issue, in this paper we first propose a Semi-Stacked Denoising Autoencoders (Semi-SDAE) model; then a new hybrid CF model incorporating the proposed Semi-SDAE model into matrix factorization, HCF-SS model, is developed. The HCF-SS model can flexibly use various sources of side information and the recommendation accuracy is improved. Experiments on two real-world datasets demonstrate that the proposed HCF-SS model outperforms the compared models.