Oktavia Putri Handayani, Purwono Purwono, Alfian Ma'arif, Iswanto Suwarno
Early detection of breast cancer is a critical challenge in medicine because it strongly influences treatment success and patient survival. Although machine learning has been widely applied to breast cancer detection, issues such as class imbalance and suboptimal hyperparameter selection often degrade model performance. This study proposes an optimized Light Gradient Boosting Machine (LightGBM) combined with Adaptive Synthetic Sampling (ADASYN) for class rebalancing and calibration-aware evaluation to improve predictive reliability on the Wisconsin Diagnostic Breast Cancer (WDBC) dataset. Hyperparameters are tuned with Optuna, and the resulting model is compared against a Support Vector Machine (SVM) baseline. Evaluation uses 10-fold cross-validation with Accuracy, Precision, Recall, Fl-score, and training/inference time. The optimized LightGBM with ADASYN attains 99.1% accuracy and an Fl-score of 0.991, outperforming the SVM (98.0% accuracy and 0.974 Fl-score) while providing more efficient inference. Model interpretability is assessed with SHapley Additive explanations (SHAP), which identify worst concave points and mean radius as the most influential features for classification, supporting clinical plausibility. Overall, the approach demonstrates that a tuned gradient-boosting model can deliver robust, efficient, and interpretable medical classification and may contribute to transparent clinical decision-support systems. © 2025 IEEE.
Department of Informatics, Universitas Harapan, Bangsa, Purwokerto, Indonesia; Department of Electrical Engineering, Universitas Ahmad Dahlan, Bantul, Indonesia; Department of Electrical Engineering, Universitas Muhammadiyah, Bantul, Yogyakarta, Indonesia
Research at a Glance
Register to unlock
Topics & SDG Alignment
Register to unlock
Collaboration
Register to unlock
Author Profile (Selected)
Register to unlock
References Overview
Register to unlock
Journal & Source
Register to unlock
Metadata & Integrity
Register to unlock