Welcome to gingado!

[!NOTE]

Archived repo

Thank you for your interest in gingado. This repo contains the first gingado codes up until the v0.2.0 release. This repo is now archived. Subsequent developments and new versions starting from v0.2.1 are now hosted at the Bank for International Settlements’ Monetary and Economic Department repo here. Users looking for the latest versions or to active contribute with code, opening issues, etc should go to that repo.

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 more data from official sources, improving the machine models being trained by the user;

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

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

• (new!) 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!

[!TIP]

New functionalities are planned over time, so consider checking frequently on gingado for the latest toolsets.

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.

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 nbdev.

Presentations, talks, papers

The most current version of the paper describing gingado is here. The paper and other material about gingado (ie, slide decks, papers) in this dedicated repository. Interested users are welcome to visit the repository and comment on the drafts or slide decks, preferably by opening an issue. I also store in this repository suggestions I receive as issues, so users can see what others commented (anonymously unless requested) and comment along as well!

Install

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

$ pip install gingado

If you use this package in your work, please cite it as below:

Araujo, Douglas KG (2023): “gingado: a machine learning library focused on economics and finance”, BIS Working Paper No 1122.

@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}
}

References

<div id="refs" class="references csl-bib-body hanging-indent" entry-spacing="0">