The travel industry also uses predictive analytics in a number of other capacities as listed below:
Dynamic offers of travel products
Travellers are increasingly searching for personalised experiences. Without predictive analytic this task is impossible. In few milliseconds systems need to predict what the customers wants and then adapt the offer accordingly. For example bringing together a flight, extra baggage, a hotel, while also slightly adapting the price dynamically and considering competitors offers in real-time.
Segmentation and clustering of passengers
A basic segmentation of travellers is based on their trip purpose (e.g., leisure or business). This is a well-known driver to explain customer behaviour (e.g., price sensitive vs product quality) and then used to adapt travel products accordingly. Supervised classification algorithms can be used to segment people in well-known classes. On the other hand, clustering algorithms permit discovery of new kinds of behaviours; the ones that cannot be classified by hand easily. This is a key feature to segment customers from new generations and dissimilar cultures.
There are more than 3 billion air passengers per year worldwide. This generates a large number of online transactions that need to be validated in real-time. The losses coming from fraud in payments and other related transactions are significant. Predictive analytics play an important role detecting frauds and even cyber-attacks on travel systems. The main challenge on these kinds of applications is to balance the trade-off between the number of “true” detected issues against the “false positive” cases. You may know cases of credit card rejections when traveling abroad, well…this is a “false positive” that should be taken into account to improve these algorithms.
Passenger and other travel data enrichment
Analysis of records coming from different travel systems have shown that data is not as rich as we would like: many database fields are not commonly filled (e.g., large presence of missing values). Similarly there is other important information that is not coming from raw data and are available only by integrating other data sources. Predictive analytics is useful to infer missing data and also matching different sources bringing new capabilities to travel systems.
Forecasting (eg., for revenue management systems and demand analysis)
Historically, forecasting engines embedded in revenue management systems considered only past bookings. However, what happens if some elements are changed? (e.g., a new “fare family”), what happens if an airline move their flight from 9:00AM to 8:00AM? What is the impact on demand if the price of flight tickets increases 10%? New generation forecasting systems are based on “factors” (price, schedules, etc.) in addition to simply time-series, thus allowing them to optimise revenue.
(Article originally published by amadeus blog)