A Modified Brain Storm Optimization Based Feature Selection For Parkinson Disease Classification

Abstract

Parkinson’s disease (PD) is an advancing neurological conditionchiefly resulting in sequential motor problems. Therapy necessitates therobust and describable diagnosis pertaining to the severity level for PD. However, very les data is available with respect toserious PD patients, however there is massive amount of data available for moderately critical PD patients, and due to this unbalanced distribution, the diagnosis accuracy is reduced.During the early phases of the disease,PD patients primarily experience vocal impairments. Therefore, diagnosis systems that are founded on vocal conditions leadthecurrent  studies on PD detection. This study developed a feature selection approach based on modified brain storm optimisation and a classification model that uses a discrete wavelet transform (DWT) for signal modification. Features such as wavelet Shannon entropies, energies, zero-crossing rates (ZCR), Mel frequency cepstral coefficients (MFCC), and linear predictive coding (LPC) are derived. The Modified Brainstorm Optimisation, which is used to reduce the features, is then used to pick the features. Lastly, the PD data is classified using a classifier based on Support Vector Machines (SVM). The outcomes of simulation shows that the proposed technique provides the optimal accuracy resultsso that treatment and therapy of PD patients are empowered.