서포트 벡터 회귀를 이용한 프로야구 팀의 승률 예측: 타자 경기력 유형 조합에 따른 설명 가능성을 중심으로*
Predicting Baseball Team Win Rates Using Support Vector Regression: Focusing on the Explainability of Batter Performance Type Combinations
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
본 연구의 목적은 프로야구 타자의 경기력 유형 상호작용조합 변수와 투수 지표를 활용하여 팀의 시즌 누적 승률을추정하고, 선형성과 비선형성을 고려한 입력 특성(feature)의 중요도를 평가하는 것이다. 이를 위해 본 연구에서는2021~2023시즌 MLB 팀의 경기 자료를 수집하여 회귀문제를 다루는 서포트 벡터 머신(SVR; Support Vector Regression)을 통해 분석을 수행하였다. SVR에서는 선형 커널과 비선형 커널을 모두 적용하였으며, 각 모형에서투입된 특성 변수가 팀의 승률에 미치는 영향을 추정하기 위해 선형 커널에서는 추정된 계수를 고려하였고, Radial Basis Function(RBF), Polynomial 커널에 대한 결과에서는 Shapley Additive Explanations(SHAP) 분석을 통해팀의 승률에 영향을 미치는 주요 변수를 탐색하였다. 연구 결과, 모든 SVR 모델에서 장타 기반의 타점 생산 능력이좋은 타자와 장타 기반의 하위타선 유형(cluster 2*cluster 3)의 상호작용조합이 팀의 승률에 미치는 영향력이 가장큰 것으로 나타났다. 또한 추정된 승률에 대한 손실함수 평균 제곱 오차(MSE; Mean Square Error)는 linear 커널에서가장 낮게 나타났으나, 실제 값과 예측값을 시각화하여 살펴본 결과 비선형 커널을 활용한 예측에서 일부 특잇값이MSE에 영향을 미치고 있는 것으로 나타났다. 본 연구의 결과는 스포츠 데이터에 내재된 비선형성(nonlinearity)을고려하는 것에서 나아가, 특정 변수에 초점을 맞추어 설명변수들의 상호작용이 결과변수에 미치는 비선형적 영향을확인적 접근(confirmatory approach)을 통해 분석하였다는 점에서 의의가 있을 것이다.
Abstract
This study aims to estimate a professional baseball team’s cumulative season win rate using interaction combinations of batter performance types and pitcher indicators and to evaluate the importance of input features considering both linearity and nonlinearity. For this purpose, the study collected game data from MLB teams for the 2021-2023 seasons and analyzed using Support Vector Regression (SVR). Both linear and nonlinear kernels were applied in the SVR models. To estimate the impact of input features on team win rates, estimated coefficients were considered for the linear kernel, and Shapley Additive Explanations (SHAP) analysis was used to explore key variables affecting team win rates for the Radial Basis Function (RBF) and Polynomial kernels. The results showed that in all SVR models, the interaction combination of batters with strong slugging-based run production abilities and those with many slugging-based sacrifice hits (cluster 2 * cluster 3) had the most significant impact on team win rates. Additionally, the Mean Square Error (MSE) for the estimated win rate was lowest with the linear kernel. However, visualization of the actual and predicted values indicated that some outliers in the nonlinear kernel predictions influenced the MSE. The findings of this study are significant in that they explain professional baseball team win rates by considering both linear and nonlinear relationships, moving beyond the assumption of linearity in sports data analysis.
참고문헌 (17)
- 이기봉, "MLB 타자의 경기력 유형에 따른 팀 승률 변화 분석", 한국체육측정평가학회지, 25(4), pp.35~46 (2023)[DOI]
- 최형준, "스포츠경기분석의 이슈와 전망", 한국체육측정평가학회지, 22(3), pp.105~113 (2020)[DOI]
- Albini, E., "Counterfactual shapley additive explanations", Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency, pp.1054~1070 (2022)
- Basak, D., "Support vector regression", Neural Information Processing-Letters and Reviews, 11(10), pp.203~224 (2007)
- Benavidez, S., "Prediction of Future Offensive Performance of MLB Position Players", ~ (2019)
- Chicco, D., "The coefficient of determination R-squared is more informative than SMAPE, MAE, MAPE, MSE and RMSE in regression analysis evaluation", Peerj computer science, 7, pp.e623~ (2021)
- Fort, R., "Quantile insights on market structure and worker salaries : The case of Major League Baseball", Journal of Sports Economics, 20(8), pp.1066~1087 (2019)[DOI]
- Hamilton, R. I., "Using SHAP values and machine learning to understand trends in the transient stability limit", IEEE Transactions on Power Systems (2023)[DOI]
- Hellevik, O., "Linear versus logistic regression when the dependent variable is a dichotomy", Quality & Quantity, 43, pp.59~74 (2009)[DOI]
- Lundberg, S. M., "A unified approach to interpreting model predictions", Advances in neural information processing systems, 30 (2017)
- "Marginal revenue product and salaries: Moneyball redux" (2010)[URL]
- Rochim, A. F., "Performance Comparison of Support Vector Machine Kernel Functions in Classifying COVID-19 Sentiment", 2021 4th International Seminar on Research of Information Technology and Intelligent Systems (ISRITI), pp.224~228 (2021), IEEE
- Rohmah, M. F., "Comparison four kernels of SVR to predict consumer price index", Journal of Physics: Conference Series, 1737(1) (2021)[DOI]
- Roman, I., "In-depth analysis of SVM kernel learning and its components", Neural Computing and Applications, 33(12), pp.6575~6594 (2021)[DOI]
- Valero, C. S., "Predicting Win-Loss outcomes in MLB regular season games–A comparative study using data mining methods", International Journal of Computer Science in Sport, 15(2), pp.91~112 (2016)[DOI]
- Yaohao, P., "Non-linear interactions and exchange rate prediction : Empirical evidence using support vector regression", Applied Mathematical Finance, 26(1), pp.69~100 (2019)[DOI]
- Zhou, W., "A comparative study of 11non-linear regression models highlighting autoencoder, DBN, and SVR, enhanced by SHAP importance analysis in soybean branching prediction", Scientific Reports, 14(1), pp.5905~ (2024)[DOI]