Using Machine Learning for Detecting Liquidity Risk in Banks
Description
This is a raw anonymised dataset to reproduce results on liquidity risk detection in commercial banks. Data is shared under Attribution (CC BY) license.
Files
Steps to reproduce
This is a guideline to use code and data shared under MIT license and Attribution (CC BY) to reproduce journal article results. A capsule with this code and data was certified by CodeOcean and received a Reproducibility Badge. It was published with a DOI 10.24433/CO.5061127.v1. Code Organisation: The study has three parts implemented in both Python files and Jupyter Notebook files as follows: Part 1: Data preparation: Validation, cleaning, exploration, and transformation to required dataset format (Implemented in ProgI_DataPreparation.py or ProgI_ DataPreparation.ipynb) Part 2: Factors and features identification (Implemented in ProgII_FactorsAndFeatures.py or ProgII_ FactorsAndFeatures.ipynb) Part 3: Model building and tests (Implemented in ProgIII_ModelsAndTests.py or ProgIII_ ModelsAndTests.ipynb) Configuration: Files config_unix_filesystem.py is a configuration file. There is no need to change except the parameter path_to_module if working in Google Drive. File preparation.py checks for and installs missing libraries. Guide to Run Codes: Jupyter notebook (.ipynb) can be run starting with Part I, followed by Part 2, and lastly Part 3. Alternatively, Python files can be run by executing a run.sh file. Data Organisation: Source data are in the data/original folder, static data are in data/static, and results and models will be stored in results. The remaining folders are for passing intermediate data. The output of Prog_I will be stored in folder data/final, and the output of Prog_II will be stored in data/model_input. Inquiries: rbarongo@gmail.com