Releases: alagoz/cfire
CFIRE v1.0 – A Reproducible and Interpretable Cross-Representation TSC Framework for Research
📦 CFIRE v1.0.0 – A Reproducible and Interpretable Cross-Representation TSC Framework for Research
CFIRE (Cross-Representation Feature Extraction) is a modular, interpretable, and reproducible framework designed for academic research in feature-based time series classification (TSC).
This release provides an extensible foundation for exploring diverse feature spaces across multiple signal representations, including time-domain statistics, frequency-domain coefficients, derivatives, wavelet transforms, and Hilbert-based features.
🧪 Key Research Features
- 🔁 Cross-representation pipeline: Time, frequency, DWT, FFT, derivative, and Hilbert transform support
- 🧠 Feature-rich foundation: Integrates Catch22 and TSFresh
- ⚙️ Parallelized extraction: Efficient multiprocessing for large-scale experimentation
- 📊 Classifier benchmarking: Includes ExtraTrees, XGBoost, Ridge, SVM, and more
- 🧩 Reproducibility and compatibility: Compatible with aeon and all UCR datasets
🛠 Included in v1.0
crossfire.py: Core CFIRE implementationdemo_.py: Example script for running experiments on any UCR datasetREADME.md: Setup guide, feature descriptions, usage, and citation instructions
🎯 Use CFIRE to:
- Benchmark and compare time series representations
- Evaluate interpretable, handcrafted feature sets
- Support TSC research with reproducible baselines
- Build and deploy robust models in low-resource or explainability-critical settings
📬 Contact
For academic inquiries, feedback, or collaborations, please reach out to:
📧 celal.alagoz@gmail.com
📖 Citation
If you use CFIRE in your research, please consider citing the repository.
@software{Alagoz_CFIRE_v1_0_0_2025,
author = {Alagöz, Celal},
doi = {10.5281/zenodo.15695652},
month = jun,
title = {{CFIRE v1.0.0 – A Reproducible and Interpretable Cross-Representation TSC Framework for Research}},
url = {https://github.com/alagoz/cfire},
version = {v1.0.0},
year = {2025}
}
