Clear for take off with machine learning

People's Choice Semi-Finalist

Clear for take off with machine learning made it to the People's Choice Semi-Finals (Top 25)

THE CHALLENGE: Clear for Take Off
Aeronautics

To comply with security and airline protocol, air travelers should arrive at the airport well in advance of their flight. Without realizing the probability of adverse conditions at the time of the scheduled departure, they may experience inconvenient delays at the airport. Delays can be short and relatively easy to manage, or they can cause long hours of waiting in crowded airports. Flight delays can even cause forced overnight stays at local hotels or inside the terminal. Travelers could benefit from knowing the likelihood of a delay as it could help them prepare for the wait time. Can an app be developed that predicts the impact of weather on airplane departure times?

Explanation

Objectives:

Over 28% of flight delays every year cause large amount of financial loss. This project aims to do forecasting of the possibility of flight delays according to big data processing and analysis. Users could check the possibility of the delay of specific flights through the web application we developed.

Users:

Common tickets takers: (especially business men) to be the reference for consumers to choose flights

Insurance companies: to assess the insurance of different flight delays

Airports: to achieve a better arrangement of airport capacity

Methodology:

Data processing:

Flight delay data were collected from Bureau of Transportation Statistics, Hourly weather data were collected from the National Climatic Data centre in US. To make the process simpler, specific airlines (e.g. airlines between Boise and Las vegas) were selected and relevant data sources were processed and integrated. On one hand, relevant data at the departure and arrival location were collected and processed. On the other hand, during the flight, weather data were collected from weather stations in different area the airline across in each hour.

Neutral Network processing:

Data of different weather conditions, including wind speed, thunderstorm , temperature etc., and time condition, including hour, date, month etc., were targeted as the input elements. The delay situation of the specific airline was set as the output result. The weights of different elements of different neural layers were achieved from the training of MATLAB simulation.

APP design:

The model equations from MATLAB were compiled into PHP. The app was designed with languages, including PHP, JavaScript, Html and CSS.

Resources Used
Made inLiverpool England
from the minds of
How they did it