Income estimation model

This project trains a simple income estimation model from US census data. For model details please see the model scorecard

Install

Install dependencies using pip or creating a conda environment

   # Create a new conda workspace
   conda create --name income-estimation
   # Activate the workspace
   conda activate income-estimation
   # Install requirements
   conda install --file requirements.txt

Train

Use mlflow to train the model

   mlflow run . -e main

The serialized model, scalers and encoders will be written to the model subdirectory.

Unit tests can be run using

   mlflow run . -e test

REST API

There is a simple REST API available to make model predictions. Start the api with

   uvicorn main:app

Now you can use the predict route to make predictions (see http://localhost:8000/docs for details). There is also an example python script available in api_predict.py