Sovereign 8s.is mirror of standardgalactic/MI-SVM-for-Imbalanced-Multi-Class-Datasets
Find a file
Repository files (latest commit first)
Filename Latest commit message Latest commit date
2026-03-31 10:54:29 +05:30
main.py Update main.py 2026-03-31 10:41:07 +05:30
misvm_model.py Add files via upload 2026-03-31 10:39:16 +05:30
README.md Update README.md 2026-03-31 10:54:29 +05:30
requirements.txt Create requirements.txt 2026-03-31 10:51:17 +05:30

MI-SVM: Multiclass Imbalanced Support Vector Machine

Description

MI-SVM is a multiclass classification approach designed for imbalanced datasets. It integrates Support Vector Machines with a kernel transformation technique to enhance minority class representation. The method employs pairwise class decomposition, hierarchical classification, and support vector–based boundary detection to improve classification performance.


Features

  • Handles multiclass imbalanced datasets
  • Kernel transformation for improved class separation
  • Pairwise class decomposition strategy
  • Hierarchical classification mechanism
  • Support vector–based boundary detection

Project Structure

MI-SVM/ │ ├── main.py # Script to run experiments ├── misvm_model.py # Core MI-SVM implementation ├── requirements.txt └── README.md


Requirements

  • Python 3.x
  • numpy
  • pandas
  • scikit-learn

Install dependencies: pip install -r requirements.txt


How to Run

  1. Prepare your dataset (see preprocessing requirements below)
  2. Update dataset loading in main.py
  3. Run the script:

python main.py


Data Preprocessing Requirements

This repository does not include preprocessing scripts, as preprocessing may vary depending on the dataset.

Users are expected to ensure:

  • All input features are numeric
  • No missing values are present
  • Features are scaled/normalized if required
  • Class labels are properly encoded (e.g., integers)
  • Data is split into training and testing sets

Input Format (Example)

Feature1, Feature2, Feature3, Label
0.25, 1.45, 3.10, 1
0.60, 2.30, 0.85, 2


Output

The model provides:

  • Predicted class labels
  • Performance metrics:
    • Accuracy
    • Precision
    • Recall
    • F1-score
  • Training and testing time

Reproducibility Note

Due to randomness and system-level differences, results may slightly vary from those reported in the study. However, the overall performance trends remain consistent.


Citation

If you use this work, please cite the corresponding research paper.


License

This project is intended for academic and research purposes.