Univariate Time Series Analysis Template
If you’re looking for a standard boilerplate to analyse univariate time-series data, this is for you!
In this post, I present standard steps that can be used for time series analysis, including application in forecasting. There are three steps that we will perform:
1. Check for Stationarity. Use 3 tests: ADF, KPSS and autocorrelation.
2. If signal is non-stationary, make it stationary. Use either transformation: Detrending, Differencing or Decomposing.
3. Performs forecasting. Use SARIMAX model.
I’ve also created a Jupyter Notebook to accompany this post here.
Check for Stationarity
Forecasting using SARIMAX
With SARIMAX, we don’t need to manually perform step 2 (making signal stationary), since it is automatically done by the model.