Integrating DEA with Machine Learning for Predictive Modeling in Breast Cancer

Abstract

This study proposes an integrated methodology combining Data Envelopment Analysis (DEA) with Machine Learning (ML) to enhance predictive modeling in healthcare data analysis, specifically for breast cancer datasets. The methodology begins with essential data preprocessing steps, including data cleaning, normalization, and outlier detection, to ensure the dataset's quality and consistency. After preprocessing, DEA is applied to calculate efficiency scores for Decision-Making Units (DMUs), such as hospitals or clinics, assessing their resource utilization and performance. These efficiency scores are then incorporated as a new feature into the dataset, providing additional insights into the performance of each DMU. Various ML models are trained using the augmented dataset, and their predictive accuracy is compared to models trained on the original dataset. The inclusion of DEA-derived efficiency scores is shown to improve model performance and interpretability. The results suggest that integrating DEA efficiency scores with ML models enhances the accuracy and transparency of predictions, offering a promising approach for decision-making in complex domains like healthcare. Future research could explore the application of deep learning techniques or extend this methodology to other sectors such as energy management or financial analysis.