스포츠 예측 모델 평가

.......................................


3.5 . 스포츠 예측 모델 평가

3.5.1 . 모델 성능 측정

모델 성능을 평가하기 위해 경기 결과를 
홈 승, 원정 승 및 무승부 (스포츠에 무승부 인 경우)로 분류 한 다음 
표준 분류 매트릭스를 사용하여 모델이 올바르게 식별한 경기 수를 살펴 봅니다. 
일반적으로 관찰되는 홈 어드밴티지 현상을 감안할 때 
데이터 세트의 클래스 값에 큰 불균형이 있을 가능성은 낮지만 
홈 승리에 유리한 약간의 왜곡을 볼 수 있습니다. 
이 경우 분류 정확도가 합리적인 평가 척도입니다. 
데이터가 매우 불균형한 경우에는 ROC 곡선 평가가 더 적절할 수 있습니다.



                               교차 검증의 다이어그램


............................중략



4. Conclusions

One of the vital applications in sport that requires good predictive accuracy is match result prediction. Traditionally, the results of the matches are predicted using mathematical and statistical models that are often verified by a domain expert. Due to the specific nature of match-related features to different sports, results across different studies in this application can generally not be compared directly. Despite the increasing use of ML models for sport prediction, more accurate models are needed. This is due to the high volumes of betting on sport, and for sport managers seeking useful knowledge for modelling future matching strategies. Therefore, ML seems an appropriate methodology for sport prediction since it generates predictive models that can predict match results using predefined features in a historical dataset.


This article critically analyses some recent research on sport prediction that have used ANN, and following this, we proposed a sport result prediction ‘SRP-CRISP-DM’ framework for the complex problem of sport result prediction. Moreover, challenges facing the sport prediction application were shown to pinpoint future work for scholars in this important application. Future studies concerning ML in sport result prediction research will hopefully be benefitted by this study.


출처; https://www.sciencedirect.com/science/article/pii/S2210832717301485


Post a Comment

Previous Post Next Post