Original link:tecdat.cn/?p=10016

Original source:Tuo End number according to the tribe public number

 

This is a short demonstration that you can do with this code. Use MAPA to generate forecasts.

> Mapasimple (Admissions) T +1 T +2 T +3 T +4 T +5 T +6 T +7 T +8 t+9 t+10 t+11 t+12 457438.0 446869.3 450146.7 462231.5 457512.8 467895.1 457606.0 441295.7 471611.2 454282.0 458308.0 453472.5Copy the code

This provides a simple diagram of the sequence and forecast: a detailed view of the predicted state at each time:

 
Copy the code



In this example, I also useparal = 2. Create a parallel cluster, then shut it down. This can be used if a parallel cluster is already runningparal = 1.

Estimation and prediction at different levels of time aggregation.

The first estimate of the degree of fit of the model at each time aggregation level is also provided for identificationETSVisualization of components. The second provides in – sample and out – sample predictions.



 
Copy the code



By setting it in any of the above functionsoutplot = 0To stop drawing the output. These functions have more options to set the maximum time aggregation level, the type of MAPA composition, and so on.

The first is to enforce a specific exponential smoothing model at all aggregation levels.

 
Copy the code

In this case, the non-seasonal damping trend model is fitted to the time series. Because MAPA can no longer change between models and select a simple model, the pre-selected model may have too much freedom for a given series of summary versions. In addition, if a seasonal model is selected, the non-seasonal version of the model will be fitted for any aggregation level with non-integer seasonality. Another new option is the ability to calculate the experience prediction interval. Because these require simulation predictions to perform calculations, they are a lot of work. To get 80%, 90%, 95% and 99% predictions:

 
Copy the code
> mapa (admissions, the conf. LVL = c (0.8, 0.9, 0.95, 0.99), paral = 2)Copy the code