A Python package for performing Convergent Cross Mapping (CCM) analysis on time series data.

Note: This package is optimized for Jupyter Notebooks (.ipynb). If you need to perform CCM analysis in a Python script (.py), please use CCM_analysis_Python.

You can install the package via pip:

pip install ccm_analysis_pynb

To use this package in a Jupyter Notebook, import the function and run the analysis:

from ccm_analysis_pynb import run_ccm_analysis_jupyter

df = pd.DataFrame(data)

# Run CCM analysis
result_holder = run_ccm_analysis_jupyter(df, L=100, E=2, tau=1, THRESHOLD=0.8, save_output=False)

# Access the final DataFrame of significant relationships
final_results = result_holder.result
print(final_results)

• data_input: Either a filepath (str) to tab-separated data or a pandas DataFrame.
• L: Length of time series to consider.
• E: Embedding dimension.
• tau: Time delay.
• THRESHOLD: Significance threshold for cross-map scores.
• save_output: Whether to save plots and protocol (default: True).
• output_dir: Directory to save outputs (default: current working directory).

The function returns an interactive widget for manual evaluation of relationships. Once the evaluation is completed, it generates a final heatmap and saves the results if save_output=True.

You can test the ccm_analysis package using the sample dataset provided in this repository. You can download the test dataset from the following link:

• Test Dataset (ccm_test_data.txt)

The dataset is structured as follows:

• Columns represent the variables you want to analyze.
• Rows represent the time points for each variable.

Parts of this project are based on the Convergent Cross Mapping (CCM) implementation from the following repository from Prince Javier :

• Convergent Cross Mapping GitHub Repository Prince Javier

I have utilized parts of the CCM algorithm from this repository to help analyze causality in time series data in my own project.