Scalable recommender system based on factor analysis

TL;DR

This work targets scalable recommender systems by integrating crossed random effects with random slopes and latent-factor models to handle massive, sparse rating data. It develops EM and variational EM estimation schemes for factor models under incomplete data, deriving closed-form or efficiently computable updates for latent factors, loadings, and noise variances. The methods are extended to include user and item intercepts and covariate information, enabling robust predictions when covariates are available or sparse. The practical evaluation on MovieLens 100K demonstrates that factor models with intercepts and covariates can outperform baseline methods like softImpute in predicting user-item interactions, highlighting the approach's scalability and predictive strength in real-world settings.

Abstract

Recommender systems have become crucial in the modern digital landscape, where personalized content, products, and services are essential for enhancing user experience. This paper explores statistical models for recommender systems, focusing on crossed random effects models and factor analysis. We extend the crossed random effects model to include random slopes, enabling the capture of varying covariate effects among users and items. Additionally, we investigate the use of factor analysis in recommender systems, particularly for settings with incomplete data. The paper also discusses scalable solutions using the Expectation Maximization (EM) and variational EM algorithms for parameter estimation, highlighting the application of these models to predict user-item interactions effectively.