Design and Development of an Algorithm to Detect and diagnose Parkinson’s disease

Abstract

Parkinson’s disease is a neurological disorder that affects the neurons that produce a chemical substance known as dopamine. The symptoms of this disease are tremors, stiffness all around the body, vocal impairment. If not treated early this disease might even lead to death. This disease doesn’t have a treatment, but an early prediction might help in the reduction of the progress of the disease. Detecting the disease is much more difficult as there is no quantitative test that can be conducted to detect the disease. Voice is one of the primary symptoms of this disease and therefore the features present in the voice can be extracted and can be used to train a model to detect whether the person is suffering from the disease or not. An ensemble learning voting classifier algorithm is designed with a performance accuracy of more than ninety percent and is used for the prediction of the disease using the vocal features extracted from the person. This algorithm is trained with the dataset which contains both normal as well as the affected person’s voice features. Decision tree classifier algorithm, Logistic Regression, and that of the Support Vector Machine Algorithm are used as the input for the voting classifier which is used for detecting the disease.