An effective up-sampling approach for breast cancer prediction with imbalanced data: A machine learning model-based...

Area of Science:

• Medical Informatics
• Machine Learning
• Computational Biology

Background:

• Early breast cancer detection significantly improves treatment outcomes and patient survival rates.
• Vast amounts of clinical data exist, yet a fraction is utilized for treatment decision support.
• Imbalanced datasets are a common challenge in medical machine learning, potentially hindering model accuracy.

Purpose of the Study:

• To introduce an engineered up-sampling method (ENUS) to address imbalanced data in breast cancer prediction.
• To evaluate the impact of ENUS on the predictive performance of various machine learning models.
• To identify key features and compare the efficacy of different machine learning algorithms for breast cancer risk prediction.

Main Methods:

• Development and application of an engineered up-sampling method (ENUS) for imbalanced datasets.
• Training and validation of multiple machine learning models, including XGBoost Tree, Random Forest, and Neural Networks.
• Feature importance analysis to identify critical attributes for breast cancer prediction.

Main Results:

• ENUS significantly improved balanced accuracy (3.74%), sensitivity (8.36%), and F1 score (3.83%) for imbalanced data (minority to majority class ratio < 20%).
• XGBoost Tree with ENUS achieved high performance: 97.47% balanced accuracy, 97.88% sensitivity, and 96.20% F1 score.
• Cell_Shape and Nuclei were identified as the most influential features in predicting breast cancer, corroborating existing knowledge.

Conclusions:

• The engineered up-sampling method (ENUS) is effective in enhancing machine learning model performance for breast cancer detection with imbalanced data.
• XGBoost Tree demonstrates superior predictive power for breast cancer risk assessment when combined with ENUS.
• This data-driven approach confirms the importance of Cell_Shape and Nuclei features and offers a valuable tool for healthcare practitioners.