Improving Recommendation Performance with Auxiliary Information

Abstract

In this article, to utilize available Auxiliary information (AI) for accurate recommendation, we propose a hybrid multitask learning based recommendation approach referred to as IRAI. In contrast to previous researches, IRAI tries to leverage available auxiliary information for representation vector learning for users and items respectively. In theory, IRAI generally focuses on exploiting the intricate hidden relationship within available side information via Gaussian Process method (GP), which could reduce the prediction uncertainty, overcome the data sparsity, and provide effective and efficient item recommendation. In practice, experiments conducted over two real life datasets also could demonstrate the significant superiority of IRAI, which could provide high performance in real life applications.