You covered a lot of conceptual ground in this session. Let’s revisit some of the topics:
- This session started with the two assumptions of Auto-Regressive models. They are
- Stationarity
- Auto Regression
- Further in Stationarity, you studied the importance of stationarity in prediction. You studied the ADF and KPSS tests to detect stationarity. You checked the stationarity of Airline Passenger data with these methods in Python.
- You studied the two methods to convert a non-stationary series into a stationary series. Box-Cox Transformation makes the variance constant whereas differencing helps in removing the trend in the series by making the mean constant. You also implemented these methods in Python to make the Airline Passenger data stationary.
- In the next segment, you studied the second assumption of Auto-Regressive models which is Autocorrelation. There are two Auto Regression measures Autocorrelation Function (ACF) and Partial Autocorrelation Function (PACF). ACF captures both the direct and indirect relationship of a time series variable with its lagged values whereas PACF captures only the direct relationship between a time series variable and its lagged values. You learned the Python implementation of ACF and PACF plots.
- After studying both the assumptions of Auto-Regressive models, you built the first Simple Auto-Regressive model. The Simple Auto-Regressive model predicts future observations as a linear regression of one or more past observations. You learned the method to find the parameter ‘p’ of the AR(p) model by looking at the PACF plot. You built an AR(1) model on Airline Passenger data using Python and calculated its accuracy.
- The next Auto-Regressive model you studied was the Moving Average. The Simple Moving Average model predicts future observation as a linear regression of its past errors. You learned the method to find the parameter ‘q’ of the MA(q) model by looking at the ACF plot. You built an MA(1) model on Airline Passenger data using Python and calculated its accuracy.