PhData's ML Challenge 2: House Pricing

Project solution of the PhData's Machine Learning Challenge #2. The challenge consists on deploying a regression model for house pricing through a REST API.

The project is composed by three contenarized components (src/components):

• data_ingestion: To load the house pricing dataset.
• model_training: To train the regression models.
• model_inference: To deploy the trained models through a REST API.

Installation

Clone the repository and install the required dependencies:

git clone https://github.com/mvidela31/phdata-ml-challenge.git
cd house-pricing
pip install -r requirements.txt

For development purposes also install the dependencies on requirements-dev:

pip install -r requirements-dev.txt

Usage

Run Tests

Use pre-commit to run the lintern and formatting hook scripts:

pre-commit run -all

Use pytest to run the unit tests of the project sub-packages:

pytest . -v

Deployment

To deploy the model endpoint using Docker, first build the model inference Docker image:

cd ./src/components/model_inference & docker build --tag house-pricing/model-inference:latest .

Run the builded Docker image:

docker run -p 80:80 house-pricing/model-inference

Now the server is deployed! You can send HTTPs requests to the server endpoints. See the ./notebooks/endpoint_invocation.ipynb notebook for endpoint invocation examples.

Experiments

See the ./notebooks/experiments.ipynb notebook for models experimentation and evaluation.

Contact

• Miguel Videla Araya (miguel.videla@ug.uchile.cl)