Deploy Recommendation Models with Django

The source code for this Django web service is adapted from the tutorial available at deploymachinelearning.com

This web service makes recommendation models available with REST API. Here are important properties:

• There can be several recommendation models available at the same endpoint with different versions. Furthermore, there can be many endpoint addresses defined.
• Information about requests sent to the models are stored. This can be used later for model testing and audit.
• There tests for research code and server code.
• You can run A/B tests between different versions of recommendation models.

Code Structure

In the research directory there are:

• Code for training content-based recommendation models on the MovieLens1M dataset.
• Code for simulating A/B tests.

In the backend directory there is Django application.

In the docker directory, there are dockerfiles for running the service in the container.