I Am What I Play in Moba Game: Supervised Machine Learning Approaches to Predict Gamers’ Personalities on the Basis of Their In-Game Behaviors

Abstract

Multiplayer Online Battle Arena (MOBA) games, such as Honor of Kings, are among the most popular game genres worldwide. In 2023, approximately 150 million players in China engaged with Honor of Kings daily, highlighting the global appeal of this gaming category. Understanding how MOBA gaming behavior correlates with personality traits is valuable for both academia and industry. However, prior research predominantly relied on linear models, which often suffer from limited feature diversity and small sample sizes. This study addresses these gaps by employing five supervised machine learning models—Logistic Regression, Decision Tree, Random Forest, K-Nearest Neighbor, and Support Vector Machine —to predict personality traits based on the gaming behaviors of 1,058 participants. Key features included players’ game roles, in-game positions, and skill usage. The models aimed to predict five personality dimensions: openness, agreeableness, neuroticism, extraversion, and conscientiousness. Results revealed that machine learning models could effectively predict personality traits, achieving macro F1 scores ranging from 0.61 to 0.68. Among these, the Random Forest model demonstrated the best overall performance. As a pioneering investigation into the intersection of MOBA gaming behavior and personality traits, this study provides new insights into the potential of supervised machine learning to uncover complex behavioral patterns. The findings have implications for team management, game design, personalized player experiences, and psychological research, showcasing the utility of gaming data in understanding human behavior.

Keywords: Multiplayer Online Battle Arena, Big Five personality traits, machine learning, Personality Prediction