Welcome to gingado!

gingado seeks to facilitate the use of machine learning in economic and finance use cases, while promoting good practices. This package aims to be suitable for beginners and advanced users alike. Use cases may range from simple data retrievals to experimentation with machine learning algorithms to more complex model pipelines used in production.

Overview

gingado is a free, open source library built different functionalities:

• data augmentation, to add data from official sources, improving the machine models being trained by the user;

• relevant datasets, both real and simulated, to allow for easier model development and comparison;

• automatic benchmark model, to assess candidate models against a reasonably well-performant model;

• machine learning-based estimators, to help answer questions of academic or practical importance;

• support for model documentation, to embed documentation and ethical considerations in the model development phase; and

• utilities, including tools to allow for lagging variables in a straightforward way.

Each of these functionalities builds on top of the previous one. They can be used on a stand-alone basis, together, or even as part of a larger pipeline from data input to model training to documentation!

Install

To install gingado, simply run the following code on the terminal:

$ pip install gingado

Attribution

If you use this package in your work, please consider citing Araujo (2023).

In BibTeX format:

@techreport{gingado,
    author = {Araujo, Douglas KG},
    title = {gingado: a machine learning library focused on economics and finance},
    series = {BIS Working Paper},
    type = {Working Paper},
    institution = {Bank for International Settlements},
    year = {2023},
    number = {1122}
}

Over time, new tools that are described in specific papers might be added (eg, a machine learning-based econometric estimator). Please consider citing them as well if used in your work. Specific information, if any, can be found in the documentation.

Design principles

The choices made during development of gingado derive from the following principles, in no particular order:

• flexibility: users can use gingado out of the box or build custom processes on top of it;

• compatibility: gingado works well with other widely used libraries in machine learning, such as scikit-learn and pandas; and

• responsibility: gingado facilitates and promotes model documentation, including ethical considerations, as part of the machine learning development workflow.

For more information about gingado, please read the paper.

Acknowledgements

gingado’s API is inspired on the following libraries:

• scikit-learn (Buitinck et al. 2013)

• keras (website here and also, this essay)

• fastai (Howard and Gugger 2020)

In addition, gingado is developed and maintained using quarto.

References