Momentum Prediction Method Based on Machine Learning Algorithm

This paper presents a novel method for predicting momentum in competitive settings using machine learning algorithms. The study focuses on integrating various momentum indicators into a comprehensive assessment model. We employ a hierarchical analysis method to determine the weighting coefficients of these indicators. The model includes fundamental momentum, scoring psychology, pressure on the opponent, and offensive momentum, calculated through an exponentially weighted moving average. Additionally, we implement an XGBoost regression model for trend prediction, leveraging detailed match data. The model's effectiveness is evaluated through SHAP analysis, demonstrating its interpretative capabilities and predictive accuracy. Our approach offers a robust framework for momentum prediction, applicable across various competitive domains.