An α–β-Divergence-Generalized Recommender for Highly Accurate Predictions of Missing User Preferences

To quantify user–item preferences, a recommender system (RS) commonly adopts a high-dimensional and sparse (HiDS) matrix. Such a matrix can be represented by a non-negative latent factor analysis model relying on a single latent factor (LF)-dependent, non-negative, and multiplicative update algorithm. However, existing models’ representative abilities are limited due to their specialized learning objective. To address this issue, this study proposes an $\alpha - \beta $ -divergence-generalized model that enjoys fast convergence. Its ideas are three-fold: 1) generalizing its learning objective with $\alpha - \beta $ -divergence to achieve highly accurate representation of HiDS data; 2) incorporating a generalized momentum method into parameter learning for fast convergence; and 3) implementing self-adaptation of controllable hyperparameters for excellent practicability. Empirical studies on six HiDS matrices from real RSs demonstrate that compared with state-of-the-art LF models, the proposed one achieves significant accuracy and efficiency gain to estimate huge missing data in an HiDS matrix.

I. Introduction

People are suffering from a serious problem of extracting desired information from enormous data scattering in the ever-exploding worldwide web. Recommender systems (RSs) that are able to find their favorites out of massive data are becoming increasingly important for various applications [1]–[10]. The fundamental data source of an RS is a user–item rating matrix [2], [3], [8], where each user’s preference on each item such as movies, music, and another user, is modeled according to his/her user–item usage history. With rapidly growing user and item counts, a user touches a tiny subset of items only. Hence, a rating matrix is inevitably high-dimensional and sparse (HiDS) [8], [13], [16], [50] with numerous missing entries. For instance, the Douban matrix [26] consists of 16830839 known ratings by 129490 users on 58541 items, with 99.78% of its entries missing.