Machine Learning Methods in the Classification of the Athletes Dehydration

Regular physical activity brings improvement in health and quality of life. It is associated with reduced risk of several diseases such as cardiovascular disease and type 2 diabetes. Despite all the benefits associated with regular exercise, there are some problems that athletes face up, one of which is dehydration during and after an intense session. Studies have shown that during a protocol of dehydration of three stages, rest stage before the exercise (RE), post-exercise (PE) and after hydration (PH), there are alterations in the electrocardiographic signal observed from the heart rate variability (HRV) parameters (RR-interval, SDRR, and RMSSD). Support vector machine (SVM) and k-means have been used in bioengineering for pathologies detection and classification. The aim of this research is to evaluate the classification capability of time domain HRV parameters in the detection of dehydration in a population of athletes using SVM and k-means. A three-stage dehydration protocol was implemented (RE, PE, and PH) in a database of 16 athletes, in each stage of the protocol 10 minutes electrocardiographic signal acquisition was performed, and the RR-interval, RMSSD, and SDRR were calculated. The results obtained in this work suggest that the SVM method classifies more efficiently the stages of the dehydration protocol than k-means clustering. On the other hand, the variable that best categorized the dehydration stages was the RR-interval obtained with the VMS method with accuracy, precision and recall above 0.60. The findings of this research encourage the hypothesis that dehydration could be studied from the electrocardiographic signal.