Sensor-Based Injury Prediction in Football Using a New Interval-Valued Fuzzy Inference Library

In many computer science and engineering applications, uncertainty is a common feature that must be addressed. In this paper, our primary emphasis is on forecasting lower-body injuries in football (soccer) players, caused by overtraining or undertraining, that do not involve direct physical contact. The dataset employed in the research comes from sensors worn during training and matches by players from the KKS Lech Poznan team, which often supply distorted or incomplete data. To address this issue, an uncertainty approach is necessary. Interval fuzzy set theory, particularly the interval inference model (generalized fuzzy inference), has proven to be effective in dealing with uncertainty in this context. This analysis aims to develop decision-making models for monitoring the workload of football players and predicting an increased risk of injury during the upcoming microcycle. In pursuit of this goal, we introduce a novel Python library that is both versatile and user-friendly, created to facilitate the definition, analysis, and interpretation of interval-valued fuzzy inference systems. The enhanced library allows an intuitive specification of interval-valued fuzzy sets and rules, enabling comprehensive fuzzy inference. It operates on a dataset encompassing athlete external load parameters, offering the capacity to forecast injury risk while considering uncertainties.