On March 31, 2017, two teams of our High Potential Program participated in the Transavia Data Science hackathon. Most of us only had been in the High Potential Program for just 6 months, so it was exciting to see how far we would come, compared to the other teams consisting of senior data scientists. Aim of the hackathon was to predict boarding times of selected flight numbers for the first part of the season of 2016. To achieve this, Transavia made data available of the amount of passengers, dates, amount of baggage and destination.
To be more effective, we decided to divide the High Potentials that participated, into a Python team (Anchormen1) and an R team (Anchormen2). During the short introduction, it became clear that the predictions would be judged based on the average boarding times of 2015. Approximately 9.30h, the data was available and the hackathon could start.
For both teams, it took quite some time to pre-process the data. The data was spread over many .csv files and it wasn’t always clear how the data could be correctly aggregated. Also selecting the right features from the big amount of information was a challenge.
A nice interruption during the hard work was a short tour of the Transavia hangar. In 30 minutes we were introduced to the process of periodic maintenance of a typical Transavia passenger airplane. Definitely interesting 😊.
Then it was back to work again to reach the first deadline of 15:00h to have a prediction of the boarding times for the first half of 2016 finished. It turned out it wasn’t that easy to get the predictions into the right format. Unfortunately, Anchormen1 didn’t succeed in this on time, so there was no intermediate score for this team. Anchormen2 did succeed, the score as -2% off baseline. Enough to optimize 😊.
Anchormen1 discovered that mid-2015, there was a noticeable reduction of boarding times. A very useful discovery, as it turned out to be the results of a change in policy of Transavia. Based on this discovery, the team decided to take only the data of the second half of 2015 to train their prediction model, which was decisive for the high end result.
Before the final results were announced, the first pitches started. Anchormen2 gave a very clear and detailed presentation of the findings that emerged from their model and how to interpret it, together with advice on how Transavia could act on it. This convinced the jury, so they awarded Anchormen2 with the first place for showing the best pitch. Great job!
All in all, it was an intense, instructive but also very fun day! Of course because of the good results, but also because of the great organization of Transavia. Thanks!