Enhancing breast cancer prediction through stacking ensemble and deep learning integration

Area of Science:

• Oncology
• Medical Informatics
• Artificial Intelligence

Background:

• Breast cancer is a significant global health concern, necessitating advancements in early detection.
• Machine learning and deep learning are increasingly vital for improving cancer diagnosis accuracy and efficiency.

Purpose of the Study:

• To evaluate ensemble and deep learning models for breast cancer diagnosis using stacking ensemble techniques.
• To enhance breast cancer prediction accuracy and efficiency through integrated modeling.

Main Methods:

• Evaluated ensemble methods (Random Forest, XGBoost, LightGBM, ExtraTrees, HistGradientBoosting, AdaBoost, GradientBoosting, CatBoost) for breast cancer diagnosis.
• Analyzed deep learning models (CNN, RNN, GRU, BILSTM, LSTM) as meta-predictors.
• Implemented stacking ensemble methodology integrating base predictors with a CNN meta-predictor.

Main Results:

• Convolutional Neural Network (CNN) demonstrated high accuracy and rapid training times, suitable for real-time applications.
• The stacking integration model combining LightGBM, ExtraTrees, CatBoost with CNN achieved high accuracy, F1 score, and ROC AUC.
• The proposed method significantly reduced training times compared to individual models.

Conclusions:

• The stacking ensemble model integrating ensemble methods with CNN offers a powerful tool for enhancing breast cancer diagnosis.
• This approach shows significant potential for healthcare decision support systems, improving diagnostic accuracy and efficiency.
• The findings provide valuable insights for optimizing breast cancer diagnosis and management strategies.