Challenge
Development of an adaptive ML system for flight prediction and evaluation that will be based on online flight data
CHALLENGES continued
The system should:
- anticipate possible flight disruptions such as delays, cancellations and diversions to other airports in real-time.
- classify in real-time possible causes of flight disruptions after their occurrence to verify the possibility of compensation EC 261/2004.
- estimate the value of a disruption claim and propose optimal handling procedures.
Unusual challenges:
- quality of aeronautical data (e.g. frequent changes to flight schedule)
- multiple data sources required
- difficulty in determining the objective reason for the delay
- many “corner cases,” e.g. strikes, changes in legislation
- verification rules change over time
Organization profile
Project implemented in cooperation with GIVT Sp. z o.o. This Warsaw-based company specializes in helping passengers obtain compensation for disrupted flights from appropriate airlines. This project is to help scale up activities that should enable expansion throughout the European market. The project will solve one of the key problems in air traffic and its disruptions.
Under EU law, airplane passengers have the right to be compensated if their flight is sufficiently delayed. The system, predicting the course of several million flights, will not only help passengers plan their trip but will allow GIVT to effectively represent its clients in claiming compensation from airlines under the aforementioned regulation.
Solution
This project aims to build the first reliable flight disruption prediction and assessment system that will use the full spectrum of available data.
Other similar systems that are currently being developed rely on limited data and therefore produce results of unacceptable significance. One of the most important aspects of this project is the creation of understandable ML systems,
as a detailed understanding of each claim is required when communicating with the airlines.
Other similar systems that are currently being developed rely on limited data and therefore produce results of unacceptable significance. One of the most important aspects of this project is the creation of understandable ML systems, as a detailed understanding of each claim is required when communicating with the airlines.
Results:
The claim is handled fully automatically in over 60 percent of cases. When detecting delayed flights, we were able to achieve a precision of 90 percent with an accuracy of 57 percent.
