Machine Learning Service

This package is used to run machine learning algorithms on query results.

Running the service

To run this service, a new deployment must be made for each algorithm, which can then be accessed using RabbitMQ. The queuename for the consumer is stored as a variable.

Dependencies

• Pika (RabbitMQ)
• Networkx

Adding a new algorithm

To add a new machine learning algorithm a new folder must be added. This file must contain a wrapper.py, Dockerfile, deployment.yml and finally a python file. The python file must implement the MLServerInterface and every function and variable it contains.

Debugging

To debug the program the user must have a way to send RabbitMQ messages. Run the wrapper.py for the desired algorithm and it will start listening to and handling messages.