Latent Factor Modeling with Expert Network for Multi-Behavior Recommendation

A team of researchers has introduced a novel AI architecture, MBLFE, designed to solve a core problem in modern recommendation systems: data sparsity. Traditional models that focus on a single user action, like a final purchase, often lack sufficient data. MBLFE instead leverages multiple user behaviors—such as clicks, views, adds-to-cart, and purchases—to build a richer understanding. The key innovation is a gating expert network that disentangles these mixed behavioral signals. Instead of creating one blurred user representation, the model trains specialized 'expert' sub-networks, each tasked with modeling a distinct latent factor or user intent.

During training, a gating network learns to dynamically select and combine the most relevant experts for each individual user. This allows the system to provide a more precise and personalized representation than holistic approaches. To ensure the experts learn truly independent factors, the researchers incorporated self-supervised learning techniques. Extensive testing on three real-world datasets demonstrated that MBLFE significantly outperforms existing state-of-the-art recommendation baselines. This validates the effectiveness of its factor-disentangling approach for capturing the nuanced motivations behind different user actions.