Football Match Analysis and Prediction Based on LightGBM Decision Algorithm

Abstract

The advent of the digital age has created new opportunities for the development of the sports industry, especially with data mining technology promoting the informatization process of the sports industry. However, there are many factors that influence football matches, and predicting their results is extremely difficult. Therefore, firstly, a dataset is constructed using a crawler algorithm and processed through various data processing techniques. Then, an improved algorithm combining the random forest algorithm and the gradient boosting decision tree algorithm is proposed. Finally, a fuzzy grey relational analysis method is designed by combining fuzzy theory and grayscale correlation model. The research results indicated that in the two groups before and after performing feature engineering operations, although the feature quantity decreased by 48.8% after the operation, the accuracy and area under the curve of the improved algorithm were the highest, with 95.31% and 86.74%, 0.9124 and 0.9767, respectively. In comparison with other mainstream algorithms, the fusion improvement algorithm and fuzzy grey relational analysis method had the highest accuracy, F1 value, and area under the curve, corresponding to 97.26%, 93.71%, and 0.9885, which were 0.12% and 0.06% higher than the accuracy of all features and area under the curve results, respectively. The above results indicate that the proposed method has superior analytical and predictive performance, which can further explore effective information, providing an effective analytical and predictive method for football related personnel and enterprises.