Collaborative Filtering Algorithm Based on SDAE and Time Mean Model

摘要

Aiming at solving the problem of data sparsity and ignorance of user interest drift in traditional collaborative filtering recommendation, this paper propose a novel collaborative filtering algorithm by combing the automatic feature extraction capability of stacked denoising autoencoder (SDAE) and the ability of real-time recommendation and rapid processing of big data of time mean model (TMM). Time factor of ratings are fully considered, and the proposed model can effectively alleviate the information distortion in feature extraction of traditional recommendation algorithms. Experimental results show that SDAE-TMM can greatly improve the performance of real-time recommendation system.