CoPilot

Inspiration

Rationalizing and optimizing the work of professionals is our passion! Even small improvements can have profound impacts on productivity and

Finnair already has a great internal dashboard for increasing situational awareness of their planners. However, figuring out which specific flights currently need attention could be much easier. We wanted to make it immediately obvious which flights are currently most risky.

What it does

CoPilot is a web application that ranks flights based on their risk of being delayed. This risk is calculated using machine learning with weather, social feed and internal data, and flight schedules. The current dashboard that Finnair uses to track risk factors shows very general information. Our application allows the user to track risk factors and other useful information on a per flight basis.

How we built it

For the machine learning classifier that calculates the risk, we needed to create our own data set by combining several sources. Shortly put, we took the risk labels, or values, from the dashboard data file and matched the labels with weather data based on the time. An SVM classifier was trained with the data set using sk-learn and python.

The user interface is a web application built using modern frontend technologies, including React.js, Redux, Typescript and RxJS. The design focus was on informative and dynamic data visualization that serves professionals using the dashboard in their day-to-day work.

Challenges we ran into

Creating the data set for our classifier was somewhat challenging and created several problems. For example, the data files were large, which slowed down the process, and matching the two tables together based on the time required creating a custom Excel script that wasn't quite straightforward.