Parkinson’s Disease Classification Using Fuzzy-Based Optimization Approach And Deep Learning Classifier

Abstract

Limited care is provided to PD (Parkinson’s disease) affected individuals due to inadequate, irregular monitoring of symptoms, occasional care taken, light involvement of clinicians that leads to reduced effective decision and sub-optimal patient health-based results. In the starting period of PD, individuals commonly have vocal impairments. Hence, vocal problem based diagnosis method was the foremost research for PD. The irrelevant and/or redundant features are eliminated in feature selection method. These chosen features provide the best result using the objective function. For most of the cases, it is a NP-hard (Nondeterministic Polynomial-time hard) problem. From last 5 years, the database size has been increased and hence there is need for feature selection before performing any classification method. To solve this problem, Fuzzy Monarch Butterfly Optimization Algorithm (FMBOA) feature selection algorithm is introduced in this work. This algorithm selects most important features from the dataset and increases the PD detection rate. Firstly, KPCA (Kernel based Principal Component Analysis) dimensionality method is introduced for reducing dimension in the dataset. Secondly FBOA based feature selection; weight value is the essential factor that is used for searching optimal features in the PD classification. In the proposed FMBOA algorithm, weight value is computed via the Gaussian fuzzy membership function. A new event is performed in the proposed Fuzzy Monarch Butterfly Optimization Algorithm where the weight value of Butterfly Optimization Algorithm is modified while performing the optimization process to enhance the results. The classification algorithms are used for varied feature set that are obtained from ABOA and each set have different combinations. The FCBi-LSTM (Fuzzy Convolution Bi-Directional Long Short-Term Memory) is developed for PD classification. The introduced framework was evaluated using UCI repository of machine learning and LOPO CV is used for performance validation. The measures that are considered for performance evaluation are MCC, f-measure and accuracy.