Integration of multi-objective PSO based feature selection and node centrality for medical datasets.

Area of Science:

• Computer Science
• Medical Informatics
• Bioinformatics

Background:

• The proliferation of large-scale medical datasets necessitates advanced computational methods.
• High-dimensional medical data presents challenges in speed and accuracy for diagnostic applications.
• Feature selection is a key dimensionality reduction technique for improving diagnostic models.

Purpose of the Study:

• To propose a novel Particle Swarm Optimization (PSO)-based multi-objective feature selection method.
• To enhance the efficiency and effectiveness of disease diagnosis using feature selection.
• To address the computational complexity associated with high-dimensional medical datasets.

Main Methods:

• A three-phase approach was developed for feature selection.
• Phase 1: Graph representation model for original features.
• Phase 2: Calculation of feature centralities within the graph.
• Phase 3: An improved PSO-based search for final feature selection.

Main Results:

• The proposed method was evaluated on five diverse medical datasets.
• Demonstrated improvements in efficiency compared to existing methods.
• Showcased enhanced effectiveness in feature selection for medical applications.

Conclusions:

• The novel PSO-based feature selection method offers significant advantages for medical data analysis.
• The approach effectively balances accuracy and computational complexity.
• This method represents a promising advancement in medical informatics and bioinformatics.