Optimization-Based Ensemble Feature Selection Algorithm and Deep Learning Classifier for Parkinson's Disease.

Area of Science:

• Computational neuroscience
• Medical informatics
• Machine learning

Background:

• Parkinson's Disease (PD) is a debilitating neurodegenerative disorder with significant impact on aging populations.
• Accurate and early diagnosis of PD is critical for improving patient outcomes and survival rates.
• Data mining techniques (DMTs) and machine learning techniques (MLTs) offer promising avenues for enhancing PD diagnosis.

Purpose of the Study:

• To develop and evaluate an advanced machine learning model for precise early detection of Parkinson's Disease.
• To investigate the efficacy of optimization-based ensemble feature selections (OBEFSs) in identifying optimal feature subsets for PD classification.
• To integrate multiple feature selection algorithms, including FMBOAs, LFCSAs, and AFAs, within an ensemble framework.

Main Methods:

• Utilized optimization-based ensemble feature selections (OBEFSs) to combine multiple feature selection algorithms (FMBOAs, LFCSAs, AFAs).
• Developed a classification model using fuzzy convolution bi-directional long short-term memories (FCBi-LSTMs) trained on OBEFS-selected features.
• Employed the University of California-Irvine (UCI) learning repository and Leave-One-Person-Out-Cross-Validations (LOPO-CVs) for model evaluation.

Main Results:

• The proposed OBEFSs method successfully identified optimal feature subsets for PD classification.
• The FCBi-LSTMs model, utilizing OBEFS-selected features, demonstrated high accuracy in PD detection.
• Evaluations using accuracy, F-measure, and Matthews correlation coefficients (MCCs) confirmed the model's robust performance.

Conclusions:

• The integration of OBEFSs with FCBi-LSTMs provides a powerful and accurate approach for early Parkinson's Disease detection.
• Ensemble feature selection methods enhance classification performance by leveraging the strengths of diverse algorithms.
• This study highlights the potential of advanced machine learning techniques in addressing critical challenges in neurological disorder diagnosis.