Early Detection of Parkinson's Disease from Sleep Efficiency using Optimized Machine Learning Models

Abstract

Parkinson's disease (PD) a type of neurodegenerative disorder that affects elderly people. Patients with Parkinson's disease mostly have symptoms of muscle rigidity, speech difficulty, movement challenges. PD when not detected at the beginning may cause severe health issues, thus premature detection of PD is crucial and more challenging. The major early symptom for PD is sleep disorder, frequent awakening in the night, this leads them to lesser the quality of sleep, resulting in disease worsening as the time increases. With the sleep efficiency the PD can be detected at early stages when checked with other diagnosing methods including gait recognition, speech tremor data. PD can be treated through early detection; this enables patients to lead a normal life with proper medication. The rise of an aging population around the world is the major cause for the disease to be identified premature and accurately. This project, sleep efficiency data is used to identify Parkinson's disease. Machine learning models like AdaBoost (ADB), Extreme Gradient Boosting (XGB) and Extra Trees (ET) for the prediction of Parkinson's disease from sleep efficiency. Experimental results demonstrated that the AdaBoost model showed the highest accuracy of Parkinson's disease classification around 99.11%.