Developing a Predictive Model Using Machine Learning Algorithms (Python) to Estimate Muscle Injury Risks Based on Daily Training Loads

Abstract

This applied predictive study aims to develop an intelligent model using machine learning algorithms (Python) to estimate muscle injury risks in athletes based on combined daily training load metrics (internal physiological and external mechanical loads). The study utilized a dataset comprising (20) training records containing the following variables: session duration, intensity, Rated Perceived Exertion (RPE), heart rate, total distance, and sprints. The Random Forest classifier was implemented to build the predictive framework, and the dataset was partitioned into an 80% training set and 20% isolated testing set using stratified sampling to ensure statistical balance. The model evaluation results demonstrated absolute predictive efficiency and a superior capacity to separate training workload patterns. The algorithm achieved an overall Accuracy, Precision, and Recall of 100.00% on the testing set, with zero false alarms or hidden critical cases, Furthermore, feature importance analysis revealed that total distance and session duration were the most critical indicators influencing injury prediction, sharing the highest relative weight of (0.2400 each), followed closely by heart rate at (0.2000). The study concludes with a strong recommendation for sports clubs to adopt machine learning models as daily early-warning systems to support proactive decision-making and dynamically adjust training dosages, thereby mitigating muscle injury occurrences before they manifest on the field.