Abstract
With the advancement of medical technology, a large amount of complex data on cancers is produced for diagnosing and treating cancers. However, not all this data is useful, as many features are redundant or irrelevant, which can reduce the accuracy of machine learning models. Metaheuristic algorithms have been employed to select features to address this issue. Although the efficacy of these algorithms has been demonstrated, challenges related to scalability and efficiency persist when handling large medical datasets. In this study, a binary version of the Advanced Al-Biruni Earth Radius (bABER) algorithm is proposed for the intelligent removal of unnecessary data and identifying the most essential features for cancer detection. Unlike traditional methods that rely on a single approach, bABER is evaluated using seven medical datasets and compared with eight widely used binary metaheuristic algorithms, including bSC, bPSO, bWAO, bGWO, bMVO, bSBO, bFA, and bGA. Statistical tests such as ANOVA and the Wilcoxon signed-rank test are conducted to ensure a thorough performance assessment. The results indicate that the bABER algorithm significantly outperforms other methods, making it a valuable tool for improving cancer diagnosis. By refining feature selection, this approach enhances existing machine learning models, leading to more accurate and reliable medical predictions. This study contributes to improved data-driven decision-making in healthcare, bringing the field closer to faster and more precise cancer detection.