A Multiple Feature Selection based Parkinson’s Disease Diagnostic System Using Deep Learning Neural Network Classifier

Abstract

Parkinson’s disease is one among the most common neurological disorder that is increasing day by day. Parkinson’s disease could become more dangerous in not treated well in time. Hence, early and automated detection of Parkinson’s disease is greatly essential. In current research, five classifiers such as DT, NB, NN, RF and SVM are exploited for PD detection. Nonetheless, more time is greatly necessitated for using further classifiers for PD detection. Enhanced Convolutional Neural Network (ECNN) method is greatly suggested for mitigating this issue in PD detection. Primarily, input PD dataset processing is done as preprocessing steps for validating dataset quality to accomplish the process. On the basis of Multiple Feature Evaluation Approach (MFEA), feature assessment processing is done which comprises quite a lot of feature assessment and ranking algorithms to weight features worth and features set extraction. Furthermore, a feature selection model by principal component analysis is suggested for classifier accuracy improvement. To conclude, Enhanced Convolutional Neural Network (ECNN) training is done over these certain features for PD detection, besides its performance evaluation is done through various metrics.