AFM AutoInt Intelligent Recommendation System based on Attention Mechanism and Automatic Interaction Modeling

Abstract

With the rapid growth of Internet data, intelligent recommendation systems are crucial for enhancing user experience and platform efficiency. Traditional algorithms struggle with high-dimensional sparse data and complex feature interactions. To address this, we propose the AFM-AutoInt model, integrating deep learning, attention mechanisms, and automatic feature interaction modeling. It utilizes embedding layers for dimensionality reduction, attention mechanisms for adaptive learning, and multi-layer self-attention for capturing high-order interactions. Experimental results show that AFM-AutoInt outperforms traditional methods in accuracy and robustness, making it a promising solution for next-generation recommendation systems.