Machine Learning Models for Predicting Breast Cancer Recurrence: A Comparison of Tree-Based Classifiers

Background and Aims: Breast cancer is one of the most prevalent cancers among women globally, accounting for nearly one-third of new cancer cases in women and resulting in the deaths of approximately ۴۲,۰۰۰ patients each year, making it a significant public health concern. When a patient is diagnosed with breast cancer, the malignant mass must be removed, followed by predicting the disease's progression—a process this study aims to automate using machine learning methods. Methods: To assess model performance and generalizability, we used ten-fold cross-validation on the Wisconsin Prognostic Breast Cancer dataset, which contains Fine Needle Aspirate test records of ۱۹۸ cases (۱۵۱ non-recurrence, ۴۷ recurrence). Principal Component Analysis was applied to reduce dimensionality. We evaluated three tree-based classifiers: a standalone Decision Tree and two ensemble models. The ensemble models included a bagging-based Random Forest and a boosting-based model, Extreme Gradient Boosting. Model efficacy was measured using accuracy, F۱ score, and area under the Receiver Operating Characteristic curve. Results: Extreme Gradient Boosting demonstrated the highest performance across all metrics. Using ten-fold cross-validation, the mean test accuracies for Extreme Gradient Boosting, Random Forest, and Decision Tree models were ۰.۹۶۰ ± ۰.۰۲۵, ۰.۹۵۱ ± ۰.۰۳۳, and ۰.۹۴۴ ± ۰.۰۳۵, respectively. Data imbalance had minimal impact on model performance, as evidenced by F۱ scores of ۰.۹۴۵ ± ۰.۰۳۶ (Extreme Gradient Boosting), ۰.۹۳۳ ± ۰.۰۴۴ (Random Forest), and ۰.۹۲۵ ± ۰.۰۴۳ (Decision Tree). All models demonstrated strong class discrimination, with area under the Receiver Operating Characteristic curve values of ۰.۹۵۴ ± ۰.۰۳۱ (Extreme Gradient Boosting), ۰.۹۴۶ ± ۰.۰۳۶ (Decision Tree), and ۰.۹۳۹ ± ۰.۰۳۴ (Random Forest). Conclusion: This study demonstrates that machine learning models, particularly Extreme Gradient Boosting, effectively predict breast cancer recurrence. Extreme Gradient Boosting outperformed Random Forest and Decision Tree models in accuracy, F۱ score, and