Peer to peer federated learning in recommendation systems

Abstract

Recommendation systems play an important role in personalising user experiences by anticipating preferences and suggesting related products. The goal of the project is to improve recommendation systems’ effectiveness and privacy by integrating federated learning approaches. Federated learning allows model training on user devices without centralizing sensitive data. The research starts with a thorough analysis of current models for recommendation systems, emphasising content-based and collaborative filtering techniques. This serves as a foundation for understanding the strengths and limitations of conventional systems. The project contributes to the evolving field of recommendation systems by providing insights into the potential advantages of federated learning. The findings aim to address concerns related to user privacy, data security, and model personalization. From a business perspective, recommendation systems offer significant monetization opportunities. In e-commerce and content streaming platforms, well-executed recommendations can translate to increased sales and consumption. By showcasing products or content that align with users’ preferences, platforms can capitalize on these oppor- tunities and drive revenue growth.