Indian Journal of Science and Technology

Abstract

Objectives: (1) To apply various traditional classification tools, (2) To check effectiveness of the classifiers to the Parkinson Dataset (3) To use boosting classification tools and (4) Compare performance of all used classification tools and find the best accuracy classifier algorithm. Thus, the main aim of the study is to discriminate healthy people from those with PD. Methods: The methodology of this study is categorised into three stages:(1) Preprocessing and feature selection; (2) Application of classifiers; (3) Comparative study. We have used secondary dataset of voice recordings originally collected by University of Oxford by Max Little. In first step, the voice data of PD patients is collected for analysis. Then the collected data is normalized using min-max normalization followed by feature extraction. Thus, uses classification Data Mining Techniques viz., KNN, Logistic Regression, Decision Tree, SVM, Random Forest and boosting algorithm etc. to predict whether the person is healthy or has Parkinson’s disease. Finally, comparative analysis is made based on the accuracy provided by different data mining models. Findings: Results of our study reveals that GB algorithm is more accurate as compared with other models. It gives the highest accuracy, so that we recommend this algorithm to deal similar kind of studies in the future. These models are very useful in better and exact medical diagnosis and decision making. It is also found that, proposed methods are fully computerized and produce enhanced performance hence can be recommended for similar studies. Here, it is observed that Gradient Boost algorithm provide the best accuracy (100% for training and 92.02% for testing, 98.46% overall). Novelty: We have used boosting classification model for the classification of Parkinson’s disease. Our proposed method is one such good example giving faster and more accurate results for the classification of Parkinson’s disease patients with excellent accuracy. We have also compared the results with other existing approaches like linear discriminant analysis, support vector machine, K-nearest neighbour, decision tree, classification and regression trees, random forest, linear regression, logistic regression and Naive Bayes, but our proposed techniques were superior to existing studies in which gradient boost algorithm yielded an accuracy of 98.46%, so our method can be used as an effective means of computer-aided diagnosis of PD, and has important practical value.

Keywords: Data Mining, Parkinson's Disease, Classification, Boosting Algorithms, Feature Selection