Employing Data Mining Techniques on Biomedical Voice Measurements to Predict Parkinson’s Disease

Presentation author(s)

Prince John ’22, Bhopal, India


Parkinson’s disease is a degenerative brain disorder which usually presents symptoms like stiffness, body tremors, and difficulty maintaining balance and coordination. The symptoms get worse with age and usually start developing around 60 years of age.

There is currently no way to diagnose Parkinson’s using blood or laboratory tests for non-genetic cases[1]. Diagnoses are done using patients’ medical history and neurological exams. It is important to diagnose Parkinson’s early, due to very similar symptoms it has to other diseases which require different treatments.

In this project, I employ data mining and machine learning techniques to analyse the Parkinson’s data set generously donated by Max Little of the University of Oxford to the public domain. The data set was used to train an artificial neural network to classify the likelihood of having Parkinson’s disease based on the speech patterns.

The dataset has been created with biomedical voice measurements from 31 different people, 23 having Parkinson’s. In total it has about 200 instances of voice recordings. The final data set does not contain any actual audio recordings. Each sound sample has been processed into 22 discrete attributes such as average fundamental frequency, % jitter, and several other measures of vocal variations.

(1) “Parkinson’s Disease Information Page”. NINDS. 30 June 2016


Eyad Said

This site uses cookies to improve your experience. Read our Web Privacy Policy for more information.

Got it! ×