Flask Demo

Here I will go through a very simple example of how you can use flask to act as an endpoint for a ML application.

Setup

For this project we require a Python version >=3.6

python -m venv venv
source venv/bin/activate
pip install -r requirements

Running

To run the application it is pretty simple.

python app.py

Requests

There are two forms of request that can be processed by the application.

Train

To train a model you will need to provide:

• model_type This is the type of model you want to train. ['LogisticRegression', 'RandomForest', 'SVC', 'DecisionTree']
• X The data to train on. (From pd.DataFrame.to_dict)
• y The associated label. (From pd.DataFrame.to_dict)

This will need to be sent to the application as a json.

The app will return the associated model_id for the newly created model.

Predict

To generate the prediction from a previously trained model you will need:

• model_id The id of the model to generate the prediction.
• data The data to predict. (From pd.DataFrame.to_dict)

This will need to be sent to the application as a json.

The application will then return the predictions and the probabilities for each of the data points sent.