PREDICTIVE MODELING OF USER BEHAVIOR IN FRIENDSHIP REQUEST SYSTEMS: A GENERALIZABLE APPROACH FOR SAFER SOCIAL PLATFORMS

Abstract

Abstract: In this article, we present the identification, validation, and application of a behavioral model designed to predict user responses to friendship requests in modern social networking applications. With safety and trust becoming central concerns in online interactions, this study investigates how dynamic user behavior can be modeled using real interaction data. The model captures trust progression, caution levels, and openness using only simple, commonly available platform metrics such as interaction history, timing patterns, and acceptance delays. The model structure is kept intentionally minimal to promote generalizability and ease of deployment across different types of users and communities. Testing was conducted on data collected from a Facebook-like social platform that includes traditional friend request mechanisms. Simulation results demonstrate that the model predicts acceptance behavior with a relative error below 5%, and can support real-time optimization features such as adaptive filtering, behavioral safety triggers, and interface personalization. This modeling approach contributes to the development of predictive safety systems in digital social environments, offering a path forward for scalable, user-centric social application design.