AutoCOT - AutoEncoder Based Cooperative Training for Sparse Recommendation

Abstract

Currently, data sparseness problem caused by large amount of data has resulted in low recommendation quality of traditional recommendation algorithms. Aiming at this problem, this paper proposes an auto-encoder recommendation algorithm based on cooperative training (AutoCOT) that combines the auto-encoder framework with cooperative training (COT) model, which can not only better learn the non-linear relationship of data but alleviate the data sparseness problem, especially in large amount of user and item data. The experiments show that, on Movielen datasets, AutoCOT performs better in coverage, precision and recall rate when compares with the traditional collaborative filtering algorithms and pure auto-encoder recommendation algorithms.