Predicting Player Performance in Valorant E-Sports using Random Forest Algorithm: A Data Mining Approach for Analyzing Match and Agent Data in Virtual Environments

This study presents a data-driven approach to predict player performance in Valorant, an increasingly popular e-sport, using a Random Forest machine learning model. As e-sports continue to evolve within the metaverse, the need for strategic optimization and player selection has become critical. By analyzing a dataset containing player statistics from the Valorant Champion Tour (VCT), we aimed to predict player Rating, a key performance indicator. The dataset includes various metrics such as Kills Per Round, Average Combat Score (ACS), Clutch Success Ratio, and Kills:Deaths. After preprocessing the data, which involved handling missing values and feature engineering, the dataset was split into training and testing sets (80% and 20%, respectively). The Random Forest model, with 100 estimators and a maximum depth of 10, was trained on the processed data. The model's performance was evaluated using regression metrics such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-squared (R²). The results demonstrated that the model could predict player performance with a high degree of accuracy, with an R² value of 0.8831, meaning it explained 88.31% of the variance in player ratings. Furthermore, Kills Per Round emerged as the most significant feature, followed by Kill, Assist, Trade, Survive Ratio and Average Damage Per Round. These insights suggest that key metrics like kills and damage output are crucial for predicting player success. This study not only provides a comprehensive framework for predicting Valorant player performance but also demonstrates the potential of data mining in optimizing e-sports strategies. The findings contribute to the growing body of research on virtual gaming environments and offer actionable insights for teams in the metaverse, enabling data-driven decision-making to enhance performance and strategic outcomes.