Intelligent medical disease diagnosis using improved hybrid genetic algorithm--multilayer perceptron network

Area of Science:

• Computational intelligence
• Machine learning in medicine
• Bioinformatics

Background:

• Genetic algorithms (GAs) are optimization techniques inspired by natural selection.
• Existing GAs can lose valuable information during the optimization process.
• Multi-layer perceptron networks require careful parameter tuning and feature selection for effective medical diagnosis.

Purpose of the Study:

• To introduce an improved genetic algorithm (GA) procedure for enhanced optimization.
• To develop a novel crossover technique, Segmented Multi-chromosome Crossover, for better information inheritance.
• To apply the improved GA for automatic and simultaneous parameter optimization and feature selection in medical disease diagnosis using multi-layer perceptron networks.

Main Methods:

• The study proposes an improved GA based on the principle that fitter parents produce healthier offspring.
• A new crossover technique, Segmented Multi-chromosome Crossover, is introduced to preserve gene segment information and facilitate multi-parent inheritance.
• The enhanced GA is applied to optimize parameters and select features for multi-layer perceptron networks in medical datasets.

Main Results:

• The improved GA demonstrated superior performance in medical disease diagnosis compared to previous methods.
• It achieved the highest average accuracy for diabetes and heart disease datasets.
• The algorithm secured the second-best average accuracy for the cancer dataset.

Conclusions:

• The proposed improved genetic algorithm effectively optimizes parameters and selects features for medical diagnosis.
• The Segmented Multi-chromosome Crossover technique enhances information preservation and inheritance.
• The enhanced GA offers a promising approach for improving the accuracy of diagnostic systems in healthcare.