Flight delay prediction system - Solution to the aeronautic Challenge from NASA Space Apps Challenge 2016 Clear for Take Off. It was developed in less than 48hs.
The server source code in at this repository
The potential and complexity of the solution is that the prediction is not based on old statistics, but it is based on the future and current forecasts combined with aviation reports and multiple factors that influence the weather-related delays, such as the position of the runways at the airport, the number of runways, the aircraft specifications, the conditions for emergency landings at the departure airport and the International Aviation Standards and Regulations. This information is public; the key point is to combine the correct factors and to generate the most precise rules, and that is why GOT a Flight is being created!
- Implement Supervised machine learning techniques to gain even more accuracy upon each delayed flight predicted. Currently the application is hosted in IBM Bluemix, so a potential candidate could be the Watson service.
- Add more forecast sources and validate estimates: detect if the weather prediction of a particular source is not reliable at a particular time.
- Include historical statistics and compare them with each prediction to test the results and adjust the estimate if found beneficial.
- Check and analyze non-weather data to cover more delay factors, apart from weather conditions.
- Load more aircraft models and airports with the corresponding specifications.
- Insert the flight number instead of the itinerary.
- Refactor of code
- API documentation