Parkinson ’S Disease Diagnosis Using An Ensemble Deep Learning Model By Utilizing Voice Features

Abstract

A chronic and progressive neurodegenerative disease, Parkinson's disease (PD) affects speech patterns, motor control, and cognitive abilities, making early and precise diagnosis vital for effective intervention. However, many existing diagnostic models face significant limitations, such as sensitivity to outliers in normalization processes and the inability to effectively capture temporal patterns in voice data critical for early-stage PD detection. Additionally, conventional models often suffer from poor pre-process, inadequate feature selection, which can lead to overfitting and decreased performance in categorization.  To move through these difficulties, this paper proposes a novel ensemble-based diagnostic model that integrates advanced pre-processing, optimization, and classification techniques. The approach begins by applying the Synthetic Minority Oversampling Technique (SMOTE) to address class imbalance. A hybrid normalization strategy combining Z-Score and Min-Max normalization is employed to improve feature scaling while reducing the influence of outliers. For feature selection, the Improved Whale Optimization Algorithm (IWOA) is used to identify the most discriminative and relevant features. The classification process leverages a powerful ensemble of models: An Enhanced Adaptive Neuro-Fuzzy Inference System (EANFIS), a Deep Belief Network (DBN) with hyperparameters optimized using the Pelican Optimization Algorithm (POA), and a Modified Long Short-Term Memory (M-LSTM) network. This ensemble exploits the combined strengths of fuzzy logic, deep feature learning, and temporal sequence modelling. The suggested model performs noticeably better than existing techniques on a number of criteria, including accuracy, precision, recall, and F1-score, according to experimental data.  Combining efficient optimization with deep hybrid learning, and temporal modelling makes this framework a promising and effective tool for the early and reliable diagnosis of Parkinson’s Disease.