A Novel Optimization Based SqueezeNet With Dwarf Mongoose Optimization Algorithm for Feature Selection and Deep Learning Classifier for Parkinson’s Disease

Abstract

A neurological condition that affects nerve cells is called Parkinson's disease (PD). Numerous neurological disorders, such as tremors, limb stiffness, and changes in walking patterns, are brought on by PD. Preserving the health of individuals affected by PD requires early detection of the condition.  However, early PD diagnosis is challenging due to the aging population. Deep learning (DL) models have been tested for accuracy and for this test, Parkinson's dataset and feature relevance analysis has been utilized. The DNN technique is utilized to evaluate and extract pressure, grip angle, and timestamp from spiral drawing inputs. The voice data in this case is sourced from Kaggle. In this study, a novel DL model for PD detection and prediction is presented. It combines an optimization technique for dwarf mongoose with a SqueezeNet. For classifications, SqueezeNet with DMOA, trained on FCBi-LSTMs generates the best features. This study's suggested model utilizes the learning repository at the University of California, Irvine (UCI). The values of Matthews correlation coefficients (MCCs) , accuracy, and F-measure values are evaluated using Leave-One-Person-Out-Cross Validations (LOPO CVs).