BEONTRA Newsletter: July 2019
BSP5 progress: Load Forecasting
At the beginning of this year, we started the development of the new BEONTRA suite with the current working title "BSP 5". The overall goal is to create solution-based workflows and open up the existing modules to each other and the outside world. In addition, we are developing a user friendly and modern user interface which is combined with advanced machine learning intelligence to deliver a state-of-the-art product platform embracing the latest technology trends.
The first solution we are currently working on is "Load Forecasting," i.e. the creation of loads (passengers, bags, fuel etc.) on the most recent schedules. With our machine learning algorithm significantly reducing the manual effort required, the Load Forecasting platform is the centerpiece to determine a flight-by-flight passenger forecast.
Currently, we are working on the first beta-version which will be made available to test users to try out the new machine learning algorithm and its predictions. The scope of the first version will be to publish the flight schedule in B Tactical, make it available within the new Load Forecasting platform, train the machine learning model with statistical data and do a flight-by-flight forecast for the total number of passengers. After the prediction is done, the data can be fed back into B Tactical to further work with it.
To validate our initial Machine Learning Algorithm, we started to compare the customer forecasts made within B Tactical with the Machine Learning results. Both were checked against the historical data. So far, we have analyzed 15 airports and focused on the following metrics. We were very happy to conclude that the target values have been achieved for all airports.
|Metric||Target||Average Result||Best Result|
|Mean Absolute Error||<25 PAX||17 PAX||13 PAX|
|Seasonal Mean Relative Percentage Error||<7%||3,4%||2,5%|
|Seasonal Mean Absolute Percentage Error||<25%||22%||13%|
|Seasonal Mean R2 Score||>75%||83%||91%|
If you are interested in the results for your airport, please feel free to contact our Product Owner Kristin Wodzinski ([email protected]) and share your B Tactical scenarios and your historical data with her.
The capacity vs. demand challenge – BEONTRA's new Capacity Demand Dashboard
Airports are going to run out of capacity due to the phenomenal growth of the aviation industry. We are confronted with this constantly. Still, the question remains: what can your organization do to meet the rising demand in the best way? BEONTRA can support you to identify the biggest capacity constraints right now, how the situation evolves in the future and how you can react to it in order to deliver excellence to your passengers in the most efficient way. This is where the new Capacity Demand Dashboard comes into play: it supports you to make better CAPEX investment decisions and to provide argumentation for strategic airport development decisions.
The two main input parameters are future (design day) schedules on the one hand and the airport's infrastructure on the other hand. Either full seasons or peak day schedules can be utilized. Design days (usually 8th, 15th or 20th busiest days for a future season, or IATA defines the design day as the 2nd busiest day in an average week during the peak month) for each year are the most common option. The airport’s infrastructure includes current infrastructure, construction work schedule (and related down time of infrastructure) and future expansions.
- Easy to understand interpretation of large data sets trickling down to the decision-critical aspects
- All-In-One: from the high-level overview to the smallest detail in order to identify action items and be able to understand it and support decision making
- Fitting to your airport’s needs: every airport is different, graphics can be imported, areas and related infrastructure are highly configurable
BEONTRA combines software applications for forecasting and planning in the aviation industry with years of experience and gathered know-how and best practices in order to build products that support your journey to excellence and efficiency.
After having successfully conducted multiple projects in the field of capacity planning and engaged with our user community over multiple years, we have decided to complement the existing solutions by adding new functionality to successfully execute capacity studies and provide the decision support to make substantiated infrastructure investment decisions. This has been accomplished with the Capacity Demand Dashboard.
The Capacity Demand Dashboard is summarizing the relationship of the demand for your infrastructure and its capacity for up to 20 years. It enables you to identify when you are going to run out of capacity and when it is time to expand which part of the infrastructure. The Capacity Demand Dashboard delivers the results in a straightforward manner that is suitable to be presented to the decision makers, upper management and other relevant stakeholders. Colour coding aids interpretation of complex data results.
Unleash the power of your data / Security Forecasting with BEONTRA has just become more accurate – Show Up Profiles Per Flight
A first version of an interface has been finalized that allows the usage of show up profiles for each individual flight. Besides the higher level of detail that can now be covered resulting in more accuracy of your Security Forecast, a huge advantage of the interface is the automation of updates of your show up profiles which results in more accurate security forecasts.
The frequent availability of boarding card scan data or other data sets on single flight level was the main driver and basis for the development of this feature.
- More detailed show up profiles – Single flight level resulting in more accurate Security Forecasts
- Automated updates of the profiles via the API
- Machine learning to continuously derive updated show up profiles from the raw boarding card scan data resulting in more accurate and customized Security Forecasts. This functionality is still in the conceptualization phase and it is expected to be released later.
BEONTRA's passenger flow simulation tool Terminal Capacity within B Capacity has been used to forecast passenger numbers at security for many years now. Together, the BEONTRA community has always driven the improvements in the accuracy of these forecasts. First and foremost, the underlying flight by flight passenger load forecast is of course the most crucial element. However, the terminal model with all its parameters and hence the accurately modelled passenger flows plays an equally important role to get the expected passenger numbers for each specific security check area over the day right.
One of the key elements of this model and a main driver for accuracy are the show up profiles — for the non-Terminal Capacity experts, these describe how and when originating departure passengers show up to the airport before their flight:
Above, it is shown how show up profiles are being defined in the system. This approach is still working very well and achieving very good results, especially when data availability is not ideal.
However, the fact that boarding card scanner data and other comprehensive data sets covering the show up behavior of passenger become available more and more readily drove the development of this new interface, which enables the system to read show up profiles for each individual day and flight. These detailed show up profiles are then used as the primary source for show up profiles. Should any flight not be covered, the profiles in the system will work as a fallback. As a result, two major improvements result in a more accurate Security Forecast for your airport. On the one hand, the more granular level of show up profiles is (based on the right data) per definition an accuracy gain. On the other hand, and equally important is the possibility to automatically update the profiles.
Using these external show up profiles (external because they are not provided by the user in the system, but automatically extracted from measurement data) is as simple as checking a box (please see screen shot below).
The prerequisite for show up profiles on a single flight level is a comprehensive data set and data source. The most prominent one is boarding card scans. This data is the source for deriving the show up profiles. On top of this, source algorithms can be applied to find similarities and patterns. We are currently conducting research on applying machine learning on top of the boarding card scan data to derive show up profiles. Once results are available, we will be sharing these and complete our offer regarding show up profiles on a single flight level.
The interface we build is a standard interface which enables any 3rd party application to connect to it. That means if your airport is already experimenting with algorithms to predict show up behavior or has already mechanism in place, we would be thrilled to connect them and see how we can improve your forecast.
As an example, our longstanding customer, Amsterdam Schiphol Airport has worked on a prediction model for show up profiles successfully. With our new interface and the usage of the detailed show up profiles in B Capacity we are now conducting validations to examine the improvement of the forecast.
If you want to know more about this topic or the outcome of the validation, please contact your Account Manager or Project Lead at BEONTRA.
23rd BEONTRA Summit
At the beginning of April, we held our annual BEONTRA Summit, hosted by our customer at Brussels Airport, the heart of Europe. With almost 40 clients and prospects attending from 21 different airports, airlines and ground handlers it was one of the largest events in the history of the BEONTRA Summit! The two-day event from April 4–5 was received very positively by our attendees with highlights being provided by a customer presentation from Auckland Airport and a completely new concept, a BarCamp — a participant-driven expert exchange with multiple parallel sessions. This unconventional format stimulates interaction and open discussion on analysis methods and aviation trends.
A special thank you goes to Brussels Airport for their dedication and support to make this a great event!
The BEONTRA Summit is our premiere annual event to keep our existing customers up to date on new features and developments in our software.
The opening session gave the attendees a compact version of the BEONTRA vision and received very positive feedback from the attendees. In addition, the team highlighted the advances in the software over the past year and gave the attendees a glimpse into the features we expect to deliver in the coming year.
The newly introduced format of a BarCamp session has been a great success. After the disappointment that it had nothing to do with drinks the community has proven once again to be full of ideas and topics to discuss. Sessions ranged from insights around how to interact with machine learning, discussions on transfer data collection to the effects of self-service bag drop on the passenger experience. With four slots in three parallel sessions, the biggest challenge was to select which interesting topics not to attend.
The 23rd BEONTRA Summit was also a chance to present awards to our customers in various categories. This year's winners were:
- Most Innovative Community Member: Oman Airports
- Forecasting Challenge (Greatest accuracy in forecasts created with BEONTRA): Birmingham Airport