---

💡 Overview

PyCCEA is an open-source package developed as part of ongoing doctoral research. It provides cooperative co-evolutionary strategies tailored for feature selection in large-scale and high-dimensional problems. The framework adopts a modular, decomposition-based approach and is intended for researchers and practitioners tackling complex feature selection tasks.

Note: PyCCEA is a work in progress. Stay tuned for improvements and new algorithm implementations.

💻 Installation

To install the PyCCEA package directly from PyPI, use the following command in a Python ≥ 3.10 environment:

pip install pyccea

Alternatively, if you want to install the latest version directly from the GitHub:

pip install git+https://github.com/pedbrgs/pyccea.git

Ensure you have pip and an active internet connection to download dependencies.

🔆 Quickstart

This quickstart demonstrates how to use the CCFSRFG1 algorithm — a CCEA variant with random feature grouping — to perform feature selection on the Wisconsin Diagnostic Breast Cancer (WDBC) dataset.

In this example, you will:

• Load the dataset using the DataLoader utility.
• Configure the dataset and algorithm from .toml files.
• Run the optimization process.

import toml
import importlib.resources
from pyccea.coevolution import CCFSRFG1
from pyccea.utils.datasets import DataLoader

# Load dataset parameters
with importlib.resources.open_text("pyccea.parameters", "dataloader.toml") as toml_file:
    data_conf = toml.load(toml_file)

# Initialize the DataLoader with the specified dataset and configuration
dataloader = DataLoader(dataset="wdbc", conf=data_conf)
# Prepare the dataset for the algorithm (e.g., preprocessing, splitting)
dataloader.get_ready()

# Load algorithm-specific parameters
with importlib.resources.open_text("pyccea.parameters", "ccfsrfg.toml") as toml_file:
    ccea_conf = toml.load(toml_file)

# Initialize the cooperative co-evolutionary algorithm
ccea = CCFSRFG1(data=dataloader, conf=ccea_conf, verbose=False)
# Start the optimization process
ccea.optimize()

The best feature subset found is stored in the attribute best_context_vector, a binary array where 1 indicates a selected feature and 0 indicates an unselected one.

📁 Custom datasets

Custom datasets are supported as long as they conform to the PyCCEA input schema (.parquet file with feature columns and a label column). To register a custom dataset at runtime, add an entry to DataLoader and execute the standard preprocessing, splitting, and normalization pipeline:

from pyccea.utils.datasets import DataLoader

# Path to your dataset in PyCCEA schema
data_path = "./custom_data.parquet"
dataset_name = "custom_data"

# Register the dataset path and task
DataLoader.DATASETS = {
    "task": "classification"  # or regression
    "file": data_path
}

# Load and prepare the dataset
dataloader = DataLoader(
    dataset_name=dataset_name,
    conf=data_conf
)
dataloader.get_ready()

If you prefer ready-to-use data, additional datasets already normalized to the PyCCEA format are available in the High-Dimensional datasets repository.

📚 Documentation

Full documentation, including a comprehensive user guide, step-by-step tutorials, an API reference, and contribution guidelines, is available at PyCCEA docs.

📜 Citation info

If you are using these codes in any way, please cite the following paper:

@article{PyCCEA,
    title = {PyCCEA: A Python package of cooperative co-evolutionary algorithms for feature selection in high-dimensional data},
    author = {Venancio, Pedro Vinicius A. B. and Batista, Lucas S.},
    journal = {Journal of Open Source Software},
    volume = {10},
    number = {112},
    pages = {8348},
    year = {2025}
}

---

📫 Contact

Please send any bug reports, questions or suggestions directly in the repository.