Which method should I choose?

For most data sets, you should select all the time-series forecasting methods in the Method Gallery. In most cases, this doesn't take much more time to calculate, and CB Predictor will pick which method best fits each set of data.

However, if you have many series of data (more than 1000) and very long series (more than 200 points per set), you can try to eliminate some of the time-series forecasting methods to speed up the forecasting. For example, if your data is seasonal or has an easily recognized trend (up or down), you can try eliminating smoothing or moving average methods. This speeds up the forecasting, but you add the risk of ruling out a method that works best with your data.

If one time series depends on one or more other time series, such as car sales depending on local median income or GNP, then you should use regression for the time series that depends on others. See your CB Predictor User Manual for more information on regression and time-series forecasting methods.