Page 156 - Building Big Data Applications

154 Building Big Data Applications

processing of this unstructured data is not Sci-Fi anymore. Machines are now able to
understand images and sounds, and in some cases, even better than the human brains.
This brings new opportunities for applications in the travel industry: from inspiration
(where to go?) to automation of reservations.
Click and conversion optimization for travel products and online advertising
campaignsd Online marketing is all about conversion; that is, the ability to sell products
or services with minimum exposure. Attracting users to an ad is not enough if customers
are not buying. New predictive algorithms could estimate conversion and help better
define travel products, better place the ads, and finally optimize advertising campaigns.
These advanced algorithms, fed with enriched travel data, become more and more able
to understand passenger needs, notably based on a combination of hundreds of specific
factors that have been found to be relevant for travelers. This opens new areas and
unlocks huge opportunities for online advertising.
Social media analysis (e.g., sentiment analysis and profiling)dMonitoring social
networks is a strategic task and it is not possible to do it manually anymore. For example,
it has been estimated that 90% of American travelers (who have a smartphones) share
photos and experiences about their travels on social networks. Similarly, millions of
travel-related reviews are shared on the internet every day. Sentiment analysis permits
the estimation of the polarity of these posts (e.g., reviews, tweets) in milliseconds. That
means knowing if they are positive, negative, and so forth. Predictive analytics have been
also been used on social networks to better known users, their interest, and needs. Just
tell me who your friends are and how you write and the algorithms will tell you who you
are (or what you look like).
Alerting and monitoringdThe travel industry generates huge volume of data. For
example Amadeus process more than 1 billion transactions per day in one of its data
centers. New aircraft have close to 6000 sensors generating more than 2 Tb per day.
Obviously this data cannot be analyzed by human beings. Using supervised machine
learning algorithms, known defects can be anticipated when a combination of factors is
observed much like how a set of symptoms helps doctors diagnose a particular disease
(with some probability). On the other hand, unsupervised learning algorithms have
helped detect anomalies to generate alerts when some data observation becomes
suspiciously rare.

Develop applications using data and agile API

Flight APIsdThe flight APIs allow you to find prices of flights from any given origin to
any destination. One of the APIs does not even require a destination, while giving you the
best prices for a variety of cities. It is ideal for inspiration applications aimed at offering
several options to the traveler. The flight APIs also provide airport information. You can
find the closest airport to a given location and even help the traveler autocomplete
destination or origin forms in your application prototype.