BTN recently hosted a virtual session with PredictX during their 2021 Business Travel Intelligence Summit Europe. Our CEO, Keesup Choe, was an integral panellist during this session, where he discussed how AI & predictive analytics can improve decision making. Keesup was joined by Aakanksha Jadhav, Foundry Team Member of Mastercard and Erik Mueller, President/CEO of Grasp Technologies.
During the session, Keesup explained the difference between generalised AI and machine learning: “AI encompasses deterministic models. And the problem with that, especially with travel data, is if new data comes in that the model hasn’t encountered before then it can’t deal with it. Machine learning is the application of probabilistic techniques based on pattern recognition.”
Keesup pointed out that travel industry leaders still have issues dealing with trouble data due to deterministic rules. “These rules don’t work on travel data because it is messy, there are no common identifiers and many special cases.”
What is an example of a travel solution AI and machine learning offer today?
“Currently, the way people interact with AI for travel is usually through booking tools or apps that allow you to interact with the AI live on your device. These apps have been personalised to recommend a scenario they think you should be taking. In the future, I’d like to see these apps alerting us to a cancelled flight, suggesting the best alternatives in real-time or even booking the alternative flight for us and directing us to the new departure gate.
What is deep learning?
Examples of deep learning include Apple’s Siri, unlocking devices using face recognition, and Tesla’s self-driving cars. Deep learning is based on neural networks and a massive amount of data. The measure of success for deep learning models is how they achieve the answer and how close they got to the correct answer.
During the session, Keesup mentioned the use of CAPTCHA (which stands for Completely Automated Public Turing test to tell Computers and Humans Apart), which is a test used in computing to determine whether or not the user is human.
“By using this test, you are actually training and helping machines to come up with datasets for deep learning models, using picture recognition and text identification.”
PredictX uses similar deep learning models for successful data wrangling. “When we have to collect 50 or 60 different datasets – of which very few have common identifiers, errors and duplications – it’s pretty much impossible to do this with deterministic systems and simple AI.”
Looking forwards, not backwards
“For the last 20 years, we’ve been living in an ‘analyse and fix’ world with companies looking back and seeing what they could have done better. We need to move into the AI world of ‘predict and prevent’. One of the ways to do this is to get data faster so that strategies can be automated and initiated as soon as something goes wrong.”
One of the ways PredictX is working towards achieving this is with the use of our new release, the Simulation Engine. This engine models data in real-time across a range of data inputs and modelling features. It ingests a dataset and uses it to model the spend across spend categories, basing this on previous trips taken per traveller and categorising by origin and destination city. Data is then analysed using the duration (days) of a journey and fed into the model, which forecasts the cost of return to travel with a high degree of precision. This allows our users to conduct detailed travel planning and accurate budgeting.
If you would like more information on our products and how they can be integrated into your business, get in touch.