Skip to content

Latest commit

 

History

History
71 lines (47 loc) · 1.95 KB

README.md

File metadata and controls

71 lines (47 loc) · 1.95 KB
Raw

Wave Telecom Customer Churn Application

This application builds a model using H2O Wave ML to predict which Telco Customers are most likely to churn and why. Shapley values and partial dependence plots allow the user to understand why the model thinks each customer will or will not churn.

Chrun App Screenshot

Running this App Locally

System Requirements

  1. Python 3.6+
  2. pip3
  3. JRE 11+ (needed to run H2O-3)
  4. NodeJS (Only needed for Run integration tests on local machine)

1. Run the Wave Server

New to H2O Wave? We recommend starting in the documentation to download and run the Wave Server on your local machine. Once the server is up and running you can easily use any Wave app.

2. Setup Your Python Environment

git clone git@github.com:h2oai/wave-apps.git
cd wave-apps/churn-risk
make setup

3. Run the App

make run

4. View the App

Point your favorite web browser to localhost:10101

Run Unit Tests

Optionally, you can run unit tests on this app

pytest # Run unit tests
pytest --cov=src --cov-report html # Run unit tests with coverage

This will generate a html report in htmlcov directory.

Run Integration Tests

Optionally, you can run integration tests on this app

  1. Go to your Wave folder downloaded in step 1 of Run app on local machine.
  2. Go to test folder inside Wave folder.
  3. Run npm install
  4. Go back to churn-risk app directory.
  5. Here I assume I have Wave downloaded in my home directory. If you have an already running wave instance,
python3 ~/wave/test/cypress.py -m src.app

else if you want to launch a new Wave instance automatically,

python3 ~/wave/test/cypress.py -m src.app -w ~/wave/waved -wd ~/wave/www