House Price Prediction Endpoint

How to run

Build the docker image

export DOCKER_BUILDKIT=1
docker build --tag house_price_endpoint:latest  -f Dockerfile .

Run the docker image

You can either run it in detached mode,

docker run -d -p 8000:80 house_price_endpoint:latest

or without detached mode,

docker run -p 8000:80 house_price_endpoint:latest

The app should be running at http://localhost:8000/

Endpoints

The app contains a single endpoint,

/predict

Make a request to the endpoint

curl -X POST http://localhost:8000/predict \
-H 'Content-Type: application/json' \
-d '{
    "features":
        {
            "avg_area_income":79545.45857431678,
            "avg_area_house_age":5.682861321615587,
            "avg_area_number_of_rooms":7.009188142792237,
            "avg_area_number_of_bedrooms":4.09,
            "area_population":23086.800502686456,
            "address":"208 Michael Ferry Apt. 674 Laurabury, NE 37010-5101"
        },
        "provider": "payment-api"
    }'

The output should be similar to,

{
  "features": {
    "avg_area_income": 79545.45857431678,
    "avg_area_house_age": 5.682861321615587,
    "avg_area_number_of_rooms": 7.009188142792237,
    "avg_area_number_of_bedrooms": 4.09,
    "area_population": 23086.800502686456,
    "address": "208 Michael Ferry Apt. 674 Laurabury, NE 37010-5101"
  },
  "provider": "payment-api",
  "output": 1244892.8
}

Testing

If you wish to run tests -

pip install -r requirements.txt
pytest

Performance

The endpoint can serve on average 85 requests/second on Intel i5-1135G, If the number of concurrent users are more than 90, the response time deteroriates, as shown by the locust test below.

anirbanpranto / house-price-endpoint