Optimizing Breast Cancer Detection: Integrating Machine Learning with Feature Selection

The development of early and precise breast cancer prediction strategies using computer-aided tools has significantly contributed to reducing mortality rates associated with the disease. Nevertheless, the feature selection approach remains a challenging task in identifying and characterizing cancers, particularly in clinical treatment contexts. This paper presents an in-depth experiment utilizing various feature selection techniques, including Genetic Algorithm, Recursive Feature Elimination (RFE), and SelectFromModel (SFM), to enhance accuracy and reduce dimensionality in the Wisconsin Diagnostic Breast Cancer (WDBC) dataset. These techniques are combined with machine learning techniques such as Support Vector Machine, Logistic Regression, and Gradient Boosting to predict breast cancer severity. The model’s performance is assessed using multiple measures, such as accuracy, F1-score, precision, recall, and area under the curve (AUC). The findings indicate that using a genetic algorithm for feature selection integrated with a gradient-boosting classifier produced superior results across all metrics. This strategy achieved an accuracy of 99.10%, a precision of 100.0%, a recall of 98.20%, an F1-score of 99.09%, and an AUC of 99.10%, surpassing the performance of features selected by RFE or SFM individually. These results highlight the critical role of the feature selection approach in optimizing model performance and increasing breast cancer detection accuracy.