AutoML GitHub Action

Easly do data analysis, train new model and get the prediction as an output using 🐻MindsDB🐻 inside GitHub Actions.

💾 Usage 🔌

Inside your workflow.yml file add:

job_name:
  runs-on: ubuntu-latest
  name: A job to train new model
  steps:
  - name: Train new model
    id: train
    uses: ZoranPandovski/automl-actions@v1.0.0
    with:
      dataset: "home_rentals.csv"
      target: "price"
      when: "{sqft: 500, rooms: 5}"

🏭 Inputs 🚢

Name	Required	Description
dataset	Yes	The path or URL to data. Can be path to the data file inside the repository or any public acessible URL as path to s3 file, raw github file, public JSON API etc.
target	Yes	The name of the variable that needs to be predicted.
when	Yes	The data that you want to make a prediction for e.g {"key" : "value", "key": "value"}
stop_training	No	Stop model training after X seconds.

🎠 Outputs 🎡

Name	Description
prediction	The predicted value

An example output that displays the predicted value, confidence and confidence interval.

👷 🚧 Roadmap 👷 🚧

📢 v1.0.0
- POC
- Accept input as dataset, target variable, query.
- Return prediction as output.
🚜 v2.0.0
- Add option for batch prediction using data in repo file or public accesible data.
- Add option for data analysis as output.
- Accept timeseries parameters as input e.g order_by, group_by.
🚚 v3.0.0
- Add option to upload a model to s3, GitHub repo, azure blob storage etc.

🏁 ⚠️ Disclaimer 🏁 ⚠️

🚦 For the GitHub Actions usage and billing check out the GitHub Site Policy.