Improving medical diagnosis reliability using Boosted C5.0 decision tree empowered by Particle Swarm Optimization

Area of Science:

• Biomedical data analysis
• Machine learning in healthcare

Background:

• Supervised classification accuracy is crucial for biomedical applications.
• Particle Swarm Optimization (PSO) combined with C4.5 decision tree (PSO+C4.5) is a common approach.
• Existing methods require further performance enhancement for complex medical data.

Purpose of the Study:

• To improve the classification accuracy of the PSO+C4.5 algorithm.
• To introduce a novel fitness function using the Boosted C5.0 decision tree for PSO.
• To evaluate the proposed method on diverse biomedical and microarray datasets.

Main Methods:

• The proposed method integrates a Boosted C5.0 decision tree as the fitness function within the Particle Swarm Optimization (PSO) framework, applied to a C4.5 decision tree classifier.
• The enhanced PSO+Boosted C5.0 classifier was implemented and evaluated on one microarray dataset and five UCI medical datasets.
• Performance was assessed using a repeated five-fold cross-validation method and compared against eight established classification algorithms.

Main Results:

• The proposed PSO+Boosted C5.0 method demonstrated improved performance over the standard PSO+C4.5 classifier.
• Experimental results indicated that the novel approach achieved higher classification accuracy compared to all eight benchmark methods evaluated.
• The Boosted C5.0 decision tree effectively enhanced the optimization process for classification tasks.

Conclusions:

• The integration of Boosted C5.0 decision tree as a fitness function significantly improves the performance of PSO-based classification in biomedical applications.
• The proposed method offers a more accurate and effective approach for supervised classification tasks in healthcare and bioinformatics.
• This research contributes a valuable enhancement to machine learning techniques for analyzing complex medical data.