why is DecisionTree view arranged in the Mining Model Viewer of Time Series?

Hello,
when I use the mining modle viewer of time series modle in BI Development,the decision tree view appears.
I don't understand the relationship between the decision tree and the time series modle.So I am puzzled to it.
Whould you please tell me the relationship between them and the use of the decision tree in time series modle?

Thanks.
Peter.C

 

For each time serie a descision tree is built, using the other time series and the historical trends etc. as input parameters. I.e. when trying to forecast the sales of one product the next month, the time serie algorithm builds a descsion tree for that product with the historical sales of that product, and the sales of all other products.

 

The second tab of the viewer displays the series with predictions.

 

I still don't understand why a descsion tree is needed in the time series modle.
Who can give me a hand?
Thanks.
Peter.C

 

Hello, Peter

As Erik mentioned, the regression formulae for time series are detected by building a decision tree, which takes as regressors various model variables (such as previous values of the current series as well as values of other correlated series). An example of such a formula:
Sales(Today) = 0.6*Sales(Today - 2) + 0.5*Sales(Today-1)+0.73

The regression tree viewer exposes the regresion formulae plus whatever non-linearity conditions detected in the series (conditions which typically translate into splits in the tree).
E.g.:
If (Sales(Today-1) > 500) Then
Sales(Today) = 1.1*Sales(Today - 2) + 0.73
Else
Sales(Today) = 1.2*Sales(Today-1) + 0.5


Therefore, basically, the first pane displays the formulae while the second pane forecasts based on those formulae.

thanks
bogdan

 

Hello,Bogdan
then,the decision tree will determine the regressors in various linearity conditions,won't it?

Peter.C
thanks.