Random Forest Classifier based detection of Parkinson’s disease

Abstract

Parkinson's disease is regarded as one of the world's most serious public health issues. Therefore, predicting this disease is very important in its earlier stage itself so that an early plan can be made by the people to take necessary treatments or actions against this dangerous disease. The minor symptoms of this disease are well known to the general public. However, the later stage symptoms are very hard to predict or detect by people around the world. Yet there is an increased number of researches being done to predict Parkinson’s disease, the non-motor symptoms preceding the motor one still remains as a myth. If a reliable and early stage can be predicted, a patient should be able to receive correct treatment in a rightperiod. Rapid Eye Movement (REM), olfactory loss, and sleep behaviour disorder are examples of non-motor symptoms. Developing Machine Learning models will be extremely beneficial in predicting this disease, and will play a critical role in stage-wise prediction. The proposed system is designed to predict all motor and non-motor features based on stage-wise classification. The Random Forest Classifier was used to classify Parkinson's patients, and 96 % accuracy was attained with the fewest voice features for diagnosing Parkinson's disease.