Parkinson’s Disease Detection by Using Maximal Information Coefficient

Abstract

In the detection of Parkinson’s disease based on acoustic features, there are a large number of irrelevant and redundant features in the dataset. Feature selection is a necessary data preprocessing method to remove irrelevant features and obtain high classification performance in pattern recognition. The significance of this work is to develop and use a novel method based on the maximal information coefficient (MIC) to measure the classification consistency between features and the decision. Firstly, the maximal information coefficient (MIC) is calculated to get the dependence between each feature and the decision. According to the correlation value, a feature sequence is obtained. Secondly, an optimal feature subset based on the maximal information coefficient (MIC) is selected by combining with a classification learning model. Finally, the designed algorithm is applied for Parkinson’s disease detection. The experimental results demonstrate that the proposed algorithm can effectively reduce the dimension of feature space, select disease-sensitive features, and achieve higher detection accuracy.