EMOO4IMD: Ensemble Multi-Objective Optimization for Imbalanced Medical Data

This repository contains the evaluation pipeline for EMOO on imbalanced medical datasets, including:

1. Heart Failure Clinical Records
2. Pima Indians Diabetes
3. Mammographic Mass

The core EMOO method is maintained separately in the original repository. This repository focuses on dataset-specific preprocessing, experiment execution, performance evaluation, and result generation.

What is included here

• dataset loading
• preprocessing
• train/test split
• calling the original EMOO optimizer
• Pareto front export
• test-set evaluation
• confusion matrix
• ROC/AUC
• CSV outputs for reproducibility

What is not included here

• the original EMOO algorithm implementation
• DEAP / NSGA-II optimizer internals
• raw dataset files

Repository structure

• emoo_bridge.py
• evaluation_utils.py
• run_heart_failure.py
• run_pima.py
• run_mammographic.py
• data/README.md

Data requirements

Raw dataset files are not included in this repository. Before running the experiments, obtain the required datasets separately and place them in the following local paths:

• ./data/heart_failure_clinical_records_dataset.csv
• ./data/diabetes.csv
• ./data/mammographic_masses.csv

Generated outputs

The scripts are designed to generate result files during execution. Depending on the experiment, outputs may be written to directories such as:

• ./results/heart_failure/
• ./results/pima/
• ./results/mammographic/

These outputs can include summary metrics, prediction files, confusion matrices, ROC plots, and Pareto-front exports.

Important setup

Before running the experiments, edit emoo_bridge.py so it imports and calls the EMOO optimizer from the original repository.