This system is a part of a master's thesis at NTNU, done by Martin Almvik and Mikael Rino Solstad. The thesis was delivered 6th of June, 2016.

The goal is to create an open-source web application (angularJS frontend and Django backend), available for everyone, that will predict where and when the flu will spread through air traffic, and visualize this over a time period. It uses tweets gathered from Flutrack.org for detecting influenza incidences, which the prediction is built upon. Visit their project on GitHub.

Furthermore, the system will use flight traffic data from BTS, and an algorithm to calculate the risk of infection in cities connected via air traffic. The algorithm is created by Rvachev and Longini in 1985, and our results correlates with the results they have found.

The API is hosted at flutrack-backend.herokuapp.com, and has two publicly available endpoints:

As a visualized interface to this API, we have created a simple single-page front-end. It's hosted at flutrack.almyy.xyz, and the open-source project can be found at GitHub.

The system is based on Python and Django, as well as some third party libraries. The steps required to install and reproduce our results is listed below.

The main source of the system is this Git repository, and everyone is free to use this project as sees fit. The project can be cloned or forked by the use of Git, using these commands:

git clone https://github.com/almyy/flutrack_backend.git

or

git clone https://github.com/almyy/flutrack_backend.git

The system uses MongoDB as datastore, and in order to test the API locally, you need MongoDB, which can be downloaded from here.

We have used Python 3.5.1, but should work on all installations over version 3.0. Install Python from here. This download also includes pip, which we will use to install third party libraries.

After you have downloaded Python and the project, you can navigate into the project directory, and install the following required libraries using pip:

pip install django djangorestframework tweepy requests pymongo xlrd

Django and Django Rest Framework are used to setup the API, Tweepy is used for utilizing the Twitter API, Requests makes it easier to make HTTP requests, Pymongo is used to communicate with our MongoDB instance, and xlrd is used to read .xls-files.

The system is dependent on external geocoding, and it is therefore necessary for you to get an API key for the Google Geocoding API. This can be obtained here. The key must be added as an environmental variable named 'GEOLOCATION_KEY'.

Now that you have everything installed, you should be able to run a local instance of the API. First you need to populate the database, so that it has some data to base the prediction on. We first run the setup.py script, which sets some necessary environment variables. If you ever want to deploy the API to a production server, the SECRET_KEY variable should be changed and hidden. In the project root folder, run the following command:

python setup.py

Then we can run populateDB.py:

python flutrack_backend/populateDB.py

This step may take a while. If you get import errors, you might have to add the project folder to the PYTHONPATH environment variable manually. This will populate the database with the required cities, airports, as well as some updated tweets from the Flutrack API.

When the database is populated and the secret key is set, the API is ready to be run. You can run the following command to start the API locally:

python manage.py runserver

This hosts the API locally on localhost:8000, and can be browsed using e.g. Postman. When running locally, the API also serves a browsable version through browsers, so you can visit the API in any browser.

Example request: GET localhost:8000/prediction