Improving the Accuracy of Ensemble Machine Learning Classification Models Using a Novel Bit-Fusion Algorithm for...

Area of Science:

• Artificial Intelligence in Healthcare
• Machine Learning for Medical Diagnosis
• Ensemble Learning Techniques

Background:

• Healthcare AI predominantly uses single classification models for disease detection, often yielding limited accuracy.
• Combining multiple classifier outputs (ensemble methods) enhances accuracy and enables big data analysis in medical AI.

Purpose of the Study:

• To propose and evaluate a novel bit fusion ensemble algorithm for minimizing classification error rates in healthcare AI.
• To enhance the accuracy of disease detection by integrating outputs from diverse base classifiers.

Main Methods:

• Implemented a bit fusion ensemble algorithm utilizing five base classifiers: k-nearest neighbor (KNN), Support Vector Machine (SVM), Multi-Layer Perceptron (MLP), Decision Tree (D.T.), and Naïve Bayesian Classifier (N.B.).
• Weighted and transformed classifier outputs (soft class vectors) into binary bits using a high-reliability threshold (δ = 0.9).
• Compared the proposed bit fusion algorithm against standard fusion approaches using average error rates on various datasets.

Main Results:

• The bit fusion algorithm demonstrated reduced error rates across multiple cancer datasets, including Leukemia (5.97%), Breast Cancer (12.6%), Lung Cancer (4.64%), Hepatitis (0%), Lymphoma (0%), and Embryonal Tumors (27.28%).
• Testing on datasets from UCI, UEA, and UCR repositories also confirmed a reduction in error rates.
• Achieved significantly lower average error rates compared to standard fusion methods.

Conclusions:

• The proposed bit fusion ensemble algorithm effectively minimizes classification errors in healthcare AI disease detection.
• This advanced ensemble technique offers a promising approach for improving diagnostic accuracy and handling big data in medical applications.
• The algorithm's performance across diverse datasets highlights its robustness and potential for real-world clinical use.