Parkinson's Disease Detection Using Machine Learning Techniques

摘要

Parkinson disease (PD) is a progressive neuro degenerative disorder that impacts more than 6 Mio. People around the world. Nonetheless, non-specialist physicians still do not have a definitive test for PD, similarly in the early stage of the diseased person where the signs may be intermittent and badly characterized. It resulted in a high rate of misdiagnosis (up to 25% among non-specialists) and many years before treatment, patients can have the disorder. A more accurate, unbiased means of early detection is required, preferably one that individuals can use in their home setting. The proposed system for predictive analytics is a mixture of clustering of K-means and a decision tree used to gain insights from patients. The problem can be addressed with reduced error rate with the application of machine learning techniques. Our proposed system also produces accurate results by combining the spiral drawing inputs of patients impacted by common and Parkinson's. From these drawings, the principal component analysis algorithm (PCA) for extraction of the feature from the spiral drawings and support vector machine is used for classification. UCI machine learning platform voice data collection in Parkinson's disease is used as feedback. Thus, our study results will show early detection of the disorder can promote the therapeutic care of the elderly and increase the chances of their life span and healthier lifestyle living peaceful life.