/cryptogpcausality

This repository contains the code for the paper "Sentiment-driven statistical causality in multimodal systems", by Ioannis Chalkiadakis, Anna Zaremba, Gareth W. Peters and Michael J. Chantler.

Primary LanguageMATLABMIT LicenseMIT

Sentiment-driven statistical causality in multimodal systems

This repository contains the code for the paper "On-chain Analytics for Sentiment-driven Statistical Causality in Cryptocurrencies", by Ioannis Chalkiadakis, Anna Zaremba, Gareth W. Peters and Michael J. Chantler.

Please cite the paper as follows:

Chalkiadakis, Ioannis, Anna Zaremba, Gareth W. Peters, and Michael J. Chantler. "On-chain analytics for sentiment-driven statistical causality in cryptocurrencies." Blockchain: Research and Applications 3, no. 2 (2022): 100063

Chalkiadakis, Ioannis and Zaremba, Anna and Peters, Gareth and Chantler, Michael John, On-chain Analytics for Sentiment-driven Statistical Causality in Cryptocurrencies: Supplementary appendix (April 3, 2021). Available at SSRN: https://ssrn.com/abstract=3818944

If you use the MatLab code for the Multiple-Output Gaussian Process Statistical Causal model, please also cite the following paper:

Zaremba, A.B., Peters, G.W. Statistical Causality for Multivariate Nonlinear Time Series via Gaussian Process Models. Methodol Comput Appl Probab (2022). https://doi.org/10.1007/s11009-022-09928-3

The available sentiment data contain the sentiment entropy index per news source, as well as the combined weighted sentiment indices that were used in the studies.

Contents of the current repository:

  1. Folder "appendix" contains the supplementary appendix with extra studies we conducted.

  2. Folder data contains the financial time-series (subfolder "finance"), the data and part of the code used in the causal analysis (subfolder "causality_data") the dictionaries (subfolder "dictionaries"), the text time-series per asset and news source (subfolder "sentiment"), and the constructed sentiment indices we used in the studies (subfolders "NLP1 - 2 - 3", each for the corresponding index as described in the paper).

  3. Folder "python" contains the Python code for constructing the sentiment indices.

  4. Folder "matlab" contains additional MatLab code for the Multiple-Output Gaussian Process Statistical Causal model

Please make sure to adjust the path names in some of the MatLab scripts if necessary.

Tested with Python v3.6.7 and the Gaussian Process Toolbox v3.5.

Copyright @ Authors April 2021