On the left is an example of monthly airline ticket bookings over an
eight-year period. It’s clear there are several complex components to
this historical data, including what appears to be a general growth in
bookings over time as well as a common repeating pattern. Using
machine learning, we can decompose the data into these components to
better understand what will happen in the future.
The growth component captures natrual growth or
decline over time. In this case study, we can see that the airline
business has grown over the last eight years and will likely continue
to do so going forward in 2016.
The cyclical component characterizes events that tend
to repeat every so often, such as annual or semi-annual events. For
the bookings data, this includes increases in revenue around holidays,
the beginning of travel seasons, and more.
The stochastic component holds whatever is left over
after removing the growth and cyclical information. Typically, the
stochastic component carries information relating to news and other
short-term events that have effects on booking revenue.