A time series is said to be stationary if the behaviour is similar in terms of statsitical properties, over time. Not this does not mean that it repeats exactly, for example if you repeatedly toss a die, the exact pattern of 1s, 2s, 3s, 4s, 5s and 6s, may be different, but each pattern is equally likely. Ther may be correlation between different moments, for example if it is wet today it may be more liely to be wet again tomorrow, but in a stationary time series the values of these correlations do not change. Knowing that a time series is stationary is heloful during data collection, as we know that watching the ame phenomenon for a long time effectvely is as good as replaying it from the start. It is even more helpful when making predictions, as models built from historical data using time series analysis or machine learning can be used to make predicitons about future behaviour.
Defined on page 318
Used on Chap. 14: page 318