A predictive analytics framework for forecasting soccer match outcomes using machine learning models

Predicting the outcome of a sports game is a favourite pastime for sports fans and researchers. The interest has intensified in recent years due to data availability, the development and successful implementation of machine learning algorithms, and the proliferation of internet gaming. This research focuses on developing a predictive analytics framework using machine learning or artificial intelligence models, as well as publicly available game results and weather data, to accurately predict outcomes of games in the English Premier League. Development efforts include experimentation using weather data and constructs such as fatigue and momentum. Ensemble techniques such as stacking or voting are also explored to improve the accuracy of basic machine learning models. The results are compared with those derived from the odds given by the major bookmakers to gauge the usefulness and potential applications in sports betting. • Developed machine-learning models to predict the outcome of soccer matches in the English Premier League. • Implemented four basic Machine learning algorithms as well as Light GBM, and Convolutional Neural Network. • Used Stacking and Voting algorithms to improve the accuracy of the predictive models. • Results showed promise compared to those derived from odds given by major bookmakers.