Machine Learning for Breast Cancer Classification: A Comparative Study of Grid Search-Optimized SVM, Random Forest, and XGBoost

The disease of Breast cancer (BC) poses a significant health challenge, as it is the most prevalent invasive cancer and the second most common cause of death among women. BC is a type of cancer that originates in the cells of the breast, with two tumor types: Benign and Malignant. Developing early detection techniques for BC recurrence constitutes a priority medical matter. Most existing techniques rely significantly on hand-crafted features and traditional models, which can be limited in terms of generalizability and performance in the case of heterogeneous or imbalanced datasets. Selecting the best machine learning (ML) approach to predict BC and its recurrence requires a comprehensive evaluation of different ML techniques. This study employs the Wisconsin Diagnostic Breast Cancer (WDBC) dataset to create a classification model utilizing Support Vector Machine (SVM) technology, which is further refined with Grid Search Cross-Validation (CV) optimization methods. The effectiveness of SVM classifiers is highly dependent on hyperparameter configuration. Alongside SVM this study utilized Random Forest (RF) and XGBoost for comparison. Among them, the SVM model optimized through GridSearchCV produced the most precise results. Notably, the model demonstrated reduced misclassification rates for tumors compared to previous approaches.