Travel-technology firm Sabre on how big-data analytics could help airlines warn travellers of potential disruptions days in advance and make travel sites better at predicting your dream destination.
Big data - the practice of tapping petabytes of data for useful business insights - holds much promise for Sabre.
The travel-technology firm manages the IT backbone of the global travel industry. Its Global Distribution System is the world's largest travel-reservation system, linking up more than 400 airlines - as well as hotels, car-rental firms and cruise lines - with 350,000 travel agents, and helping hundreds of millions of corporate and consumer travellers reach their destination.
Sabre wants to mine the data generated inside and outside its systems to gain insights that allow it to do a better job of meeting its customers' and operational needs.
For instance, with better modelling of common travel patterns and passenger behaviour Sabre could help airlines make more precise predictions about passenger demand, or by scrutinising patterns of use in its daily system logs it could anticipate spikes.
Gregg Webb, president of Sabre Travel Network, said: "There are a number of areas where the use of that data in aggregate, in terms of looking at demand fluctuation through the industry, would allow the industry to plan and operate more efficiently."
Unstructured data analysis
Big-data analytics platforms such as Hadoop are able to query petabyte-scale stores of data and are also adept at analysing unstructured data - data sitting outside databases - which makes them suited to mining the likes of social-media posts.
Sabre's R&D labs built a social-media analytics appliance that would allow airlines to spot the probable disruption of flights by hurricanes, days earlier than is currently possible.
The labs team used the appliance to predict when flights over the Caribbean were likely to cross the path of a hurricane by searching Twitter and Facebook posts for mentions of hurricanes and combining that information with weather-service forecasts.
Tests of the appliance showed that it would allow passengers to be notified about potential disruptions between 48 and 72 hours before their flight. In comparison the usual notification period, which relies on weather warnings from the National Weather Service, is about two hours.
Sabre's Webb said: "The reality is tracking social posts is easily as accurate as predictions from the weather services."
Mining external information sources
Robert Wiseman, CTO with Sabre Holdings, gave another example of how big-data analytics could one day allow travel websites to mine external information sources, such as a regular customer's social-media feed, and use those insights to improve trips they recommend to customers - for instance, through Sabre's travel site, Travelocity.
"If Travelocity knows that I'm a Sheffield Wednesday fan and knows that team are in the FA Cup final, then it knows that I'm much more likely to go ahead and book it," he said.
Sabre has bought three 0.5PB Oracle Big Data Appliances, which allow unstructured data to be loaded into Oracle Database 11g. The appliances will use Hadoop provided by Cloudera - relying on MapReduce to break the data down and Cassandra as a distributed database-management system. Sabre expects to begin using the appliances by the end of this year.
Wiseman said one of the biggest challenges is configuring the data for analysis - for instance, breaking it down using MapReduce - because the relative immaturity of the tools still means setting up a query requires a large degree of manual work and technical knowledge.
"As we get into this we realise that a lot of work that the data scientists are going to be doing really isn't the best use of their skills, so having some way of intelligently presenting this data for them would really give them a leg-up," he said.