02-03-2016 09:53 AM
Im using time series to forecast deactivations using forecast studio. I have data for two years. I also have activations data which im using as an input. so deactivations are showing decreasing trend only because our base is decreasing in essense based on out total units they r not decreasing but increasing. Any ideas on how to make this clear?
02-03-2016 02:14 PM
there are 2 ways to accomplish what you wanted:
1. you can populate the future input varaible values to the "base" value (e.g. 0). The forecasts should be the "base" forecasts. You can then test if the base forecasts are trending up or down
2. you can use UCM model (excluding ARIMA and ESM) and look at the OUTCOMPONENT dataset. Filter the OUTCOMPONENT dataset with _COMP_ = 'Filtered Level' and examine the _PREDICT_ column. This component is basically the "base foreccasts" without the influence of input variables.
02-08-2016 10:34 AM
Is there any reason/indicator which tells you that deactivations should be increasing? It would help the model identify increasing trend if you add that indicator as an independent variable to your model.