.png?width=100&height=115&name=Big%20Data%20Expo%20Vorm%20F%20(1).png)
.png?width=100&height=115&name=Big%20Data%20Expo%20Vorm%20E%20(1).png)
Applying machine learning to enable the creation of robust and efficient flight schedules at KLM
Wednesday 11:45 - 12:15
Lezingenzaal 3
Daan Mocking
Senior Data Scientist (KLM)
In January 2023, we founded a new Data Science team within KLM's department Integral Planning & Control (IP&C). IP&C supports the business with tooling and data, such that they can run operations as smoothly as possible. Our tools support the business in creating a robust and efficient schedule, as well as enabling them to effectively manage disruptions in the ever-volatile aviation environment. The responsibility of IP&C’s new Data Science team is to use statistics and modelling techniques to transform raw historical data into usable insights to improve IP&C’s optimization tools and support the business directly. Toward this purpose, the DS team developed a tool called Flint. Flint is a prediction tool for estimating flight punctuality (aka flight performance) up to six months before flight departure. Being able to accurately estimate a flight’s performance this early in the process, enables the creators of the flight schedule to apply the limited slack in the most impactful places in the schedule. This minimizes disruptions and increases the robustness of the schedule. Apart from building an accurate model, two factors were key for the successful adoption of Flint by the business: The implementation of explainable AI and providing insight into the uncertainty around the predictions. The project that the team is currently working on revolves around fuel consumption. Fuel consumption is heavily entangled with flight performance. If flights with volatile performances are scheduled too tight, minor schedule disruptions have a large ripple effect. During operation, these disruptions are frequently resolved by flying faster, to make up for lost time. Flying faster causes excessive fuel consumption and CO2 emissions. By creating a more effective schedule, many last-minute changes on the day of operation could have been avoided, thereby preventing these emissions (and associated costs). Earlier and more accurate estimations of the expected fuel consumption will help with creating such a schedule.
Felipe Ramos-Gaete
Data Science Leader (KLM)
In January 2023, we founded a new Data Science team within KLM's department Integral Planning & Control (IP&C). IP&C supports the business with tooling and data, such that they can run operations as smoothly as possible. Our tools support the business in creating a robust and efficient schedule, as well as enabling them to effectively manage disruptions in the ever-volatile aviation environment. The responsibility of IP&C’s new Data Science team is to use statistics and modelling techniques to transform raw historical data into usable insights to improve IP&C’s optimization tools and support the business directly. Toward this purpose, the DS team developed a tool called Flint. Flint is a prediction tool for estimating flight punctuality (aka flight performance) up to six months before flight departure. Being able to accurately estimate a flight’s performance this early in the process, enables the creators of the flight schedule to apply the limited slack in the most impactful places in the schedule. This minimizes disruptions and increases the robustness of the schedule. Apart from building an accurate model, two factors were key for the successful adoption of Flint by the business: The implementation of explainable AI and providing insight into the uncertainty around the predictions. The project that the team is currently working on revolves around fuel consumption. Fuel consumption is heavily entangled with flight performance. If flights with volatile performances are scheduled too tight, minor schedule disruptions have a large ripple effect. During operation, these disruptions are frequently resolved by flying faster, to make up for lost time. Flying faster causes excessive fuel consumption and CO2 emissions. By creating a more effective schedule, many last-minute changes on the day of operation could have been avoided, thereby preventing these emissions (and associated costs). Earlier and more accurate estimations of the expected fuel consumption will help with creating such a schedule.
In January 2023, we founded a new Data Science team within KLM's department Integral Planning & Control (IP&C). IP&C supports the business with tooling and data, such that they can run operations as smoothly as possible. Our tools support the business in creating a robust and efficient schedule, as well as enabling them to effectively manage disruptions in the ever-volatile aviation environment. The responsibility of IP&C’s new Data Science team is to use statistics and modelling techniques to transform raw historical data into usable insights to improve IP&C’s optimization tools and support the business directly. Toward this purpose, the DS team developed a tool called Flint. Flint is a prediction tool for estimating flight punctuality (aka flight performance) up to six months before flight departure. Being able to accurately estimate a flight’s performance this early in the process, enables the creators of the flight schedule to apply the limited slack in the most impactful places in the schedule. This minimizes disruptions and increases the robustness of the schedule. Apart from building an accurate model, two factors were key for the successful adoption of Flint by the business: The implementation of explainable AI and providing insight into the uncertainty around the predictions. The project that the team is currently working on revolves around fuel consumption. Fuel consumption is heavily entangled with flight performance. If flights with volatile performances are scheduled too tight, minor schedule disruptions have a large ripple effect. During operation, these disruptions are frequently resolved by flying faster, to make up for lost time. Flying faster causes excessive fuel consumption and CO2 emissions. By creating a more effective schedule, many last-minute changes on the day of operation could have been avoided, thereby preventing these emissions (and associated costs). Earlier and more accurate estimations of the expected fuel consumption will help with creating such a schedule.
Back to overview
Visit Data Expo
Interested in this lecture?
We believe data drives digital transformation
Unlocking the Power of Retrieval-Augmented Generation (RAG)
Digital Transformation for SMEs: 8 Benefits and Challenges
Subscribe for the newsletter
To top