Comparison of Machine Learning Techniques in the Classification of Parkinson’s Desease Sufferers

Abstract

Parkinson's disease is a progressive and relatively common neurodegenerative disorder in the central nervous system where sufferers can have difficulty moving. This disease has a high mortality rate in the world of around 9.3 million in 2021. Meanwhile, in Indonesia, it is estimated that as many as 12,980 people die every year due to Parkinson's cases. This increase in cases of death is due to the lack of information about the initial symptoms and dangers of the disease, besides it is important to know how to prevent it early. Early detection of Parkinson's disease can prevent symptoms of a certain age thereby increasing life expectancy. The existence of a computer-based system for diagnosing Parkinson's disease is called a classification system where the system applies the Machine Learning method. This study aims to compare the performance of algorithms in the classification system of people with computer diseases. In this study, it used methods in Machine Learning such as K-NN, Multi Layer Percepteron (MLP), Linear Regression and Support Vector Machine (SVM). The data set in this study was obtained using the Weka application. The dataset used was Parkinson's Disease data totaling 195 rows of data taken from the UCI Machine Learning Repository Datasets. The results of the experiment based on the four algorithms showed that the poor performance was the Multi Layer Percepteron approach to regression data with an RSME value of 0.459. Meanwhile, the k-Neural Network Algorithm is a good classification technique forParkinson's problem with an RMSE value of 0.1895.