Do you remember the last time you had people over for dinner? It is hard to know what every single guest wants. You try to make a menu that everybody could enjoy, but it’s hard to anticipate what everyone likes specifically, let alone if you have friends with food allergies. Equally, airlines have the same sort of issue when trying to understand how passengers choose when searching for flights. There might be different itineraries for the same destination, but which of them will they prefer? One way of finding out what travellers really want is through market research, but the best way is by listening to the data itself.
As an artificial intelligence (AI) researcher, I know that Amadeus has the capacity to harness data to the fullest. By analysing millions of airline bookings, including search logs and other relevant data sources, we can establish what kind of flights people have looked for, what results came up on their screens and what flights were finally booked – always taking into account privacy laws and regulations.
Our solutions, especially Amadeus Revenue Management Systems and Dynamic Pricing solutions, use current choice modelling techniques to harness all unique aspects of the data. Essentially, most of these techniques are based on statistical models, such as the multinomial logistic model, a method used to calculate the outcome of variables. For example: which degree will a university student choose given some already known aspects such as expected salary after graduation, years of study, tuition, etc.?
But our efforts do not end there, we’re always exploring new ways of helping our customers operate better. Some of our ongoing research projects look at the applications of machine learning and deep learning in travel, ranging from basic classification tasks to image recognition and driverless cars.
Furthermore, thankfully, our work has been delivering results. A significant amount of our investigation has been recognised by the research community. Amadeus is the first travel company to have a research paper on flight itineraries choice modelling accepted at the Knowledge Discovery and Data Mining 2017 conference, one of the world’s largest Data Science events. On top of that, we have shared our findings and views about the use of machine learning to improve performance of choice models at the International Choice Modelling Conference.
We’re proud of what we’ve accomplished so far, especially with our choice modelling improvements based on Deep Learning, a branch of machine learning that tries to imitate the way the human brain works by identifying patterns in digital representations of sounds, images and other data.
We’re always seeking new ways to support our customers with the latest technology advancements for their businesses, and we look forward to what the future has in store for the industry. Who knows? Maybe next time we talk about data and you happen to have people over for dinner, you might know in advance what everyone’s perfect meal is.