Data Mesh Approach to Predicting Soccer Match Outcomes with Dynamic In - Game Factors

Sajith Narayanan

Abstract: The study looks at a model to predict the final number of goals in soccer matches by breaking the games down into minutes and looking at both fixed and changing factors. The study used data from the four biggest soccer leagues from the 2018/19 to 2021/22 seasons. A Machine Learning (ML) model called a Multilayer Perceptron (MLP) was trained and compared with other models like multiple linear regression and random forest to predict how many goals would be scored. The research also studied how a coach?s decisions, like making substitutions or changing the team?s lineup, affect the final score. The results showed that the MLP model worked better than the other models, with an improvement of 1.42% over linear regression and 0.41% over random forest. Adding in factors like substitutions and changes in strategy made the predictions even better. It found that increasing the number of substitutions reduced the total goals scored by both teams.

Keywords: Machine Learning, Matches, Multilayer Perceptron, Soccer

How to Cite?: Sajith Narayanan, "Data Mesh Approach to Predicting Soccer Match Outcomes with Dynamic In - Game Factors", Volume 14 Issue 2, February 2025, International Journal of Science and Research (IJSR), Pages: 101-108, https://www.ijsr.net/getabstract.php?paperid=SR25202194426, DOI: https://dx.doi.org/10.21275/SR25202194426