Time Series Forecasting

[#1]

The Time Series Forecast platform builds a variety of different exponential smoothing models and automatically selects the with the best forecast performance. The platform is designed to forecast multiple time series.

Time Series Forecast

1. From an open JMP^® data table, select Analyze > Specialized Modeling > Time Series Forecast.
   
   
2. Select a continuous variable from Select Columns, and click Y (continuous variables have blue triangles).
   - More than one times series can be included in the Y  role and models and forecasts will be created for each. Here we illustrate with just one time series (Sales).
   
   
3. Select the variable that identifies the time periods and click Time (optional).

- Data must be sorted by time and equally spaced.

4. Click OK.

In Model Specifications you can view the recommended models or modify the modeling options.

Under Complete Specification, you can choose:

• Time periods to forecast (NAhead)
• Period for seasonality (e.g., monthly)
• Model Selection Strategy (AIC, BIC, or Forecasting Performance)
• Holdback sample and metric to assess fit (if chooseing Forecast Performance).

5. Click Run
   
   - Best fit time series model is displayed (e.g., MAA: Multiplicative Error, Additive Trend, No Dampening, Additive Seasonality).

- Time Series graph displays data, best fit model, forecasting intervals, and future forecast as defined by NAhead.

6. Choose to Save Results to the original data table or to a new data table from the red triangle at the top of the Report

Monthly Sales.jmp (Help > Sample Data Folder > Time Series)

Visit Predictive and Specialized Models > Time Series Forecast in JMP Help to learn more.