This repository contains the code of our published work in IEEE JBHI. Our main objective was to demonstrate the feasibility of the use of synthetic data to effectively train Machine Learning algorithms, prooving that it benefits classification performance most of the times.

🧬 An explainable machine learning pipeline predicting 5-year breast cancer survival using the METABRIC genomic dataset. It leverages XGBoost, ADASYN for medical data imbalance, and SHAP (Explainable AI) to translate complex RNA and clinical features into transparent, biologically interpretable predictions. 🏥📊

Data Science Case Study

The computing scripts associated with our paper entitled "Oversampling Highly Imbalanced Indoor Positioning Data using Deep Generative Models".

Implementation of augmentation, SMOTE, ADASYN, AE, DGAN for oversampling image data used in detection (YOLO and others one stage detectors). Maritime flags/code scenario.

Benchmarking 5 class-imbalance strategies (incl. a custom UMCE ensemble) across 3 classifiers and 12 KEEL datasets, with two-way ANOVA + Tukey HSD.

Diversity-guided composite pool oversampling with majority-only Tomek Links cleaning for imbalanced classification.

Diversity-Guided Pool Oversampling for imbalanced classification

Dealing with Imbalanced Datasets in Binary Classification: A Comparative Study of Resampling and Ensemble Methods

Streamlining the process to generate synthetic data. Just focus on the data, not the code!

Official implementation of BGPO: Boundary-Guided Composite Pool Oversampling for imbalanced classification.

IndoHome Telco Churn Prediction System - Comprehensive ML project with Streamlit deployment, Tableau analytics, and end-to-end data science pipeline for telecommunications customer retention

A question-answering project focused on handling imbalanced datasets through advanced oversampling techniques.