Open Source Library to Perform Entity Embeddings Using Neural Networks

2 points

7 years ago

Hey guys, I've been working as an undergrad researcher for the past year on the prediction of childbirth mortality. At this project I've developed a tool to perform entity embeddings on categorical variables using CNN with Keras. I tried pretty much to make it easy to use and flexible to most of the existent scenarios (regression, binary and multiclass classification), but if you find any other need or issue to be fixed, feel free to ask! :-)

I tried to add some cool stuff on the project, such as unit tests, code coverage with Codacy, continuous integration with Travis CI and auto deployment to PyPi and auto-generated documentation with Sphinx and ReadTheDocs, so if any of you is interested in how to setup your project to have these features, feel free to use it as a base project.

I'm also looking forward to any reviews about the source code, so any tip to improve the readability or even performance, its really welcome and well appreciated.

Github: https://github.com/bresan/entity_embeddings_categorical

PyPi: https://pypi.org/project/entity-embeddings-categorical/

Code coverage (nowadays reaching 97%): https://coveralls.io/github/bresan/entity_embeddings_categorical?branch=master

Thanks!

1 comment