ReSample

Overview

ReSample is a web application designed to automatically balance datasets for machine learning tasks. It helps address class imbalances in datasets, improving model generalization and performance. It provides a user-friendly interface to handle missing values, balance class distributions, visualize the results and export the processed dataset.

Additionally, ReSample features a recommendation model that suggests ten of the most optimal combinations of balancing methods based on the size and imbalance ratio of the uploaded dataset.

---

Live Demo

Experience the deployed with Streamlit app here:

---

Features

1. Upload or use a sample dataset.
2. Handle missing values with various strategies (drop, fiil with median/moda/mean).
3. Get recommendations about balancing methods based on dataset size and imbalance ratio.
4. Balance class distribution:
   • Oversampling:
     • Random Oversampling
     • SMOTE
     • Borderline SMOTE
     • B-SMOTE SVM
     • ADASYN
   • Undersampling:
     • Random Undersampling
     • NearMiss (version 1, 2 and 3)
     • TomekLinks
     • CNN
     • ENN
     • OSS
     • NCR
5. Visualize data before and after balancing (Pie & Bar charts).
6. Export the processed dataset.