Improved cancer detection through feature selection using the binary Al Biruni Earth radius algorithm.

Area of Science:

• Computational Biology
• Medical Informatics
• Machine Learning

Background:

• Medical technology generates vast, complex cancer data for diagnosis and treatment.
• Redundant or irrelevant features in large datasets can decrease machine learning model accuracy.
• Metaheuristic algorithms are used for feature selection, but scalability and efficiency remain challenges.

Purpose of the Study:

• To propose a binary version of the Advanced Al-Biruni Earth Radius (bABER) algorithm for intelligent data reduction and feature identification in cancer detection.
• To address the limitations of existing metaheuristic algorithms in handling large medical datasets.

Main Methods:

• A novel binary algorithm, bABER, was developed for feature selection.
• bABER was evaluated on seven diverse medical datasets.
• Performance was compared against eight established binary metaheuristic algorithms (bSC, bPSO, bWAO, bGWO, bMVO, bSBO, bFA, bGA).
• Statistical analyses, including ANOVA and Wilcoxon signed-rank tests, were performed for rigorous assessment.

Main Results:

• The bABER algorithm demonstrated statistically significant superior performance compared to all other evaluated methods.
• Effective identification of essential features and removal of unnecessary data were achieved.
• Enhanced accuracy and reliability of machine learning models for cancer prediction were observed.

Conclusions:

• The bABER algorithm is a highly effective tool for improving cancer diagnosis through optimized feature selection.
• This approach enhances the performance of existing machine learning models, leading to more precise medical predictions.
• The study contributes to advancing data-driven healthcare decision-making for faster and more accurate cancer detection.