Support Vector Machine-recursive feature elimination for the diagnosis of Parkinson disease based on speech analysis

Parkinson disease has become a serious problem in the old people. There is no precise method to diagnosis Parkinson disease now. Considering the significance and difficulty of recognizing the Parkinson disease, the measurement of samples' voices is regard as one of the best non-invasive ways to find the real patient. Support Vector Machine is one of the most effective tools to classify in machine learning, and it has been applied successfully in many areas. In this paper, we implement the SVM-recursive feature elimination which has not been used before for selecting the subset including the most important features for classification from the original features. We also implement SVM with PCA for selecting the principle components for diagnosis PD set with 22 features in order to compare. At last, we discuss the relationship between SVM-RFE and SVM with PCA specially in the experiment. The experiment illustrates that the SVM-RFE has the better performance than other methods in general.