Different Machine Learning Algorithms for Parkinson’s Disease Detection Using Speech Signals

Abstract

A neurodegenerative disorder affecting the brain's neurological, physiological, and behavioral systems is called Parkinson's disease (PD). In the early stages of the condition, modest differences make accurate identification challenging. “Bradykinesia”, or sluggish movements, is among the disease's typical symptoms. The disease's symptoms start to show up in middle age, and as people age, the severity of the condition worsens. A speech issue is one of PD's initial symptoms. According to this study, supervised classification algorithms that include feature selection and classification processes, such as light gradient boosting machine (LGBM), random forest classifier (RF), extra tree classifier (ET), gradient boosting classifier (GB), and decision tree classifier (DT), can accurately diagnose subjective diseases. This kind of approach could reduce the time and expense associated with Parkinson's disease (PD) screening, as it would require a limited set of clinical test criteria to make the diagnosis. The experimental results show that light gradient boosting machine is 91.03% accurate, random forest classifier is 90.96% accurate, extra tree classifier is 90.19% accurate, gradient boosting classifier is 90.13% accurate, extreme gradient boosting is 89.29% accurate, and decision tree classifier is 87.82% accurate; it is therefore determined that light gradient boosting machine is the most accurate one with the highest accuracy. When the obtained results were compared to those of earlier studies, it was found that the suggested work provides both equivalent and superior results.