/prediction_app

Build a production-grade machine learning application involving tox, circle-CI, Docker, and testing

Primary LanguagePython

Machine learning prediction application

CircleCI

GitHub repo size

GitHub forks

GitHub forks

Steps of usage

  1. Fork or clone this repo to your local machine.
  2. Navigate to the project directory and activate the virtualenv source dev/bin/activate
  3. Install requirements pip install -rrequirements.txt from the root directory.
  4. Navigate to the packages -> regression_model then run tox to make sure that everything is passing.

Basic Analyses

The following kaggle notebooks includes the complete basic analyses and EDA

Medium post part_1

Medium post part_2

Scaffold template - starter template

Notes to consider:

  • If you intended to use this project as a reference to build your own, you need to define your environmental variables in the circleci.
  • Tests are disabled for this project, would appreciate if anyone raise a pull request to include them.

Few things to consider

I structured the scaffold based on the OOP which seperate concerns of code.

  • processing folder conatains any scripts for data wrangeling, cleaning, or feature engineering.
  • trained_model folder contains any scripts dedicated to build model, tuning or any related scripts.
  • pipeline.py file contains all the procedures that should be done using the sklearn.pipeline
  • predict.py file dedicated for getting out the predictions
  • train_pipeline.py file dedicated to train the model, starting from downloading the dataset, split, apply pipeline...etc.

Output:

  • The resulted model is saved as a .pkl file versioned with the same version of the package.
  • Only one API end point is already deployed to Heroku, pull requests are welcome

Future work

  • Build extra API end points
  • Build frontend interface using streamlit
  • Complete unit tests.
  • Create a configuration file for travis

Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.

License

MIT