This is a repository of a full datascience project focused on a simple ML service: Deployment, unit tests, OOP, MLFlow. The goal is to predict prices of houses based on the Kings County dataset (avaliable here).

In this file you will find a notebook with the steps to find the best model.

In this file you will find the code to train the model, including the data preparation and hyperparameter tuning.

In this file is the flask code for the API to predict prices, using the model loaded from MLFlow.

First, to run the webapp locally you are going to have to clone the repository:

git clone https://github.com/kairosilveira/house_pricing_regression
cd house_pricing_regression/

Then, create and activate the virtual environment and install the dependencies(make sure eu you have python3 installed):

python -m venv venv
source venv/bin/activate 

Install dependencies

python -m pip install --upgrade pip
pip install -r requirements.txt

After installation is done, you can start the flask aplication running the app.py file

gunicorn --chdir src app:app

Now you can use the post method to make predictions using your local server adding /predict to the endpoint, here is an example using thunder:

{"data":[{ "date":"2014-07-17", "bedrooms":2, "bathrooms":1.0, "sqft_living":770, "sqft_lot":17334, "floors":1.0, "waterfront":0, "view":0, "condition":3, "grade":7, "sqft_above":770, "sqft_basement":0, "yr_built":1978, "yr_renovated":0, "lat":47.2997, "long":-122.269, "sqft_living15":1480, "sqft_lot15":17334 }] }