Autoencoder-based collaborative filtering

Abstract

Currently collaborative filtering is widely used in recommender systems. With the development of idea of deep learning, a lot of researches have been conducted to improve collaborative filtering by integrating deep learning techniques. In this research, we proposed an autoencoder based collaborative filtering method, in which pretraining and stacking mechanism is provided. The experimental study on commonly used MovieLens datasets have shown its potential and effectiveness in getting higher recall.