Główne pojęcia
Boosting algorithms improve breast cancer detection by optimizing recall metrics and using SHAP for explainability.
Streszczenie
This article explores the use of boosting algorithms like AdaBoost, XGBoost, CatBoost, and LightGBM to predict and diagnose breast cancer. The study focuses on optimizing the recall metric to reduce false negatives. By utilizing the University of California, Irvine dataset, the models were trained and tested to achieve high accuracy. The study also incorporates Optuna for hyperparameter optimization and SHAP method for model interpretability. Results show significant improvements in AUC or recall for all models, with a final AUC exceeding 99.41%.
Statystyki
Final AUC was more than 99.41% for all models.
False Negative reduced by 25% in AdaBoost.
LightGBM achieved a perfect recall of 1.0.