Prediction of competition results based on multi model optimization

• Published in: 2024 9th International Symposium on Computer and Information Processing Technology (ISCIPT) pp. 345 - 349
• Main Authors: Wang, Chongyi, Wang, Lidong
• Format: Conference Proceeding
• Language: English
• Published: IEEE 24-05-2024
• Subjects: Accuracy; Competitive forecasting; Computational modeling; Feature extraction; Games; Mathematical models; Momentum; Multi-layer perceptron model; Prediction algorithms; Predictive models; XGBoost
• DOI: 10.1109/ISCIPT61983.2024.10673288

In sports, a team or athlete may feel they have the momentum or power during a game, but this phenomenon is challenging to quantify. This paper chooses 14 features of the competition itself to quantify momentum, and then we convert "momentum" into a prediction of winning or losing a game. Five different models are used to simulate it, the best model is to use XGBoost model to quantify the momentum of the game, and then use Multi-layer Perceptron model to predict the outcome of the game. According to the weight of features in the model, we extracted the two most important factors in men's competition and women's competition separately and give corresponding suggestions. Finally, we use data from other competitions to verify our model, and the results show that the accuracy of our method is always above 55%, and it has good generalization ability and stability.