Predictive modelling links exercise dependence to associated psychological and behavioral risk factors

Area of Science:

• Sports Science
• Psychology
• Data Science

Background:

• Exercise Dependence (ED) is characterized by uncontrollable, excessive exercise negatively impacting life.
• Identifying risk factors for ED is crucial for timely intervention and prevention strategies.

Purpose of the Study:

• To utilize machine learning to identify behavioral and psychological factors contributing to Exercise Dependence risk.
• To develop a quantitative risk profile for ED assessment in sports contexts.

Main Methods:

• Employed a multi-step machine learning procedure for model construction and validation, including feature selection and bootstrapping.
• Collected data from 1099 participants across three time points (GR2021-23) using the Exercise Dependence Scale-Revised (EDS-R).

Main Results:

• Perfectionism was the most significant predictor of ED risk, followed by Drive for Thinness and Drive for Muscularity.
• The predictive models demonstrated robust performance across multiple datasets (GR2021-23), with R² values ranging from 0.58 to 0.79.
• 5.6% of participants were classified as "At Risk" of ED.

Conclusions:

• Machine learning effectively identifies key psychological and behavioral predictors of Exercise Dependence.
• The study provides a foundation for developing interpretable predictive tools for assessing ED risk.
• Understanding the interplay of personality and behavioral dimensions is vital for ED risk assessment.