Sistema de matchmaking híbrido para valorant: otimização e balanceamento com clustering k-means e classificação Random Forest

Abstract

This work presents the development of a matchmaking system for team-based first-person shooter (FPS) games, such as Valorant, using machine learning techniques to improve match balance. The main goal is to propose a hybrid solution by combining clustering algorithms (K-Means) to segment player profiles and predictive models (Random Forest) to evaluate match quality. The methodology includes collecting, preprocessing, and analyzing gameplay data extracted from Valorant. The experiments showed that the proposed model is capable of forming balanced matches by considering metrics such as player impact, diversity of roles, and tactical balance between teams. [resumo fornecido pelo autor]