상대적 성취도와 심리적 행동 패턴을 고려한 리그 오브 레전드 승패 예측 모델

초록

While the importance of data analysis in League of Legends has increased with the growth of the global esports industry, existing studies are limited by relying on simple numerical features, failing to reflect relative achievements by position and psychological behavioral patterns. This study proposed new features incorporating 'Relative Achievement' to quantify lane gaps and 'Psychological Behavioral Patterns' derived from recent match history, and we conducted win-loss prediction experiments. Experimental results using Riot API data and five machine learning algorithms showed that the proposed model achieved higher accuracy, outperforming the base model. SHAP Value analysis revealed that the relative achievement of the Bottom lane was the most critical factor across all tiers, and psychological variables had a more significant impact on match outcomes in lower tiers than in higher tiers. This study contributes to improving the accuracy and explanatory power of the prediction model by quantifying relative achievement and psychological elements.

키워드