Time series forecasting | MATLAB implements AR, ARMA, ARIMA time series forecasting model Q&A
Table of contents
basic introduction
- AR
Autoregressive Model (Autoregressive Model), often referred to as AR model, is a statistical model used for time series analysis and forecasting. It predicts future values based on the historical values of the time series itself, by modeling the relationship between the observed value at the current moment and the observed value at the previous moment.
The basic idea of the AR model is that the value at the current moment can be predicted from the value at the previous moment. Specifically, an AR§ model represents the value at the current instant as a linear combination of p instants in the past. Parameter estimation of AR models is usually performed using least squares or maximum likelihood methods. Choosing an appropriate order p is also an important issue, which can be determined by methods such as information criteria (such as AIC, BIC) or cross-validation.
Although the AR model can capture the autocorrelation relationship of the sequence, it has certain limitations, especially for the non-stationary time series modeling effect may not be good. In this case, you can combine the difference operation to make the sequence stable, or use the ARIMA model, where I means integrated, which is used to deal with non-stationarity.
In short, the AR model is one of the basic models in time series forecasting.