6 Time series decomposition

Time series data can exhibit a huge vari­ety of pat­terns and it is help­ful to cat­e­go­rize some of the pat­terns and behav­iours that can be seen in time series. It is also some­times use­ful to try to split a time series into sev­eral com­po­nents, each rep­re­sent­ing one of the under­ly­ing cat­e­gories of pattern.

In this chap­ter, we con­sider some com­mon pat­terns and meth­ods to extract the asso­ci­ated com­po­nents from a time series. Often this is done to help under­stand the time series bet­ter, but it can also be used to improve forecasts.

• 6/1 Time series components
• 6/​2 Mov­ing averages
• 6/3 Clas­si­cal decomposition
• 6/​4 X-​​12-​​ARIMA decomposition
• 6/​​5 STL decomposition
• 6/​6 Fore­cast­ing with decomposition
• 6/7 Exer­cises
• 6/8 Fur­ther reading