FEXUM

Framework for Exploring and Understanding Multivariate Correlations

This repository contains the backend of our big data framework for exploring and understanding:

Feature selection
Feature extraction
Multivariate correlations
Time series

The frontend and correlation algorithm are located in different repositories.

The basis for this application has been made into a paper, which was submitted to and accepted by ECML PKDD 2017, and published as part of the conference proceedings in the "Lecture Notes in Computer Science" (LNCS) series.

Contributors

Getting Started

Run $ docker-compose build to build all containers (do that when you are changing dependencies)
Run $ docker-compose up to start all containers. Put -d for headless mode
Run $ docker-compose run web_wsgi python3 manage.py migrate to apply migrations
Access the docker ip (mostly localhost) at port 80.

Python Shell

$ docker-compose run web_wsgi python3 manage.py shell

This is useful when you want to test and create database objects, e.g.:

>> from features.tasks import *
>> from features.models import *
>> for feature in Feature.objects.all() 
   ... a_test_name.delay(feature.name)

Task Monitoring

Open the docker host at port 5555

Running with docker in your own network environment

By default the network is configured for the docker containers as followed:

networks:
  fexum:
    ipam:
      driver: default
      config:
        - subnet: 10.151.100.0/24

If this does not work with your network setup, change the subnet.

Development & Contributing

Testing

For testing simply run:

$ docker-compose run web_wsgi py.test

Data Migrations

Create migrations after changing or creating models

$ docker-compose run web_wsgi python3 manage.py makemigrations

KDD-OpenSource/fexum