"Input Relevance" Window (ASCMO-DYNAMIC )

The "Input Relevance" window (Model > Input Relevance (RMSE) >* ) can be opened for all or a single output. The settings you define here affect the plot in the "Relevance of Inputs" Window (ASCMO-DYNAMIC) .

The Input Relevance uses model prediction of the current model to determine the influence of specific inputs on the output. Snippets of input data are exchanged at random per input and changes to the prediction are measured.

"Configuration Dynamic Input Relevance" area

"Snippet Length" input field:

Enter the number of time step into which the training data should be fragmented during the analysis. Make sure that all effects of interest are observable within this time period.

"Number of Dice Rolls" input field:

Enter the value how often the process should be repeated. The probabilistic method is repeated by the number of dice rolls/the entered value. If the amount of data is large, you can set the value to 1 to increase the performance.

"Evaluate on Dataset Categories" checkboxes

Activate the checkboxes of the dataset categories you want to evaluate.

OK

Closes the window and opens the "Relevance of Inputs" Window (ASCMO-DYNAMIC) with your settings applied.

Cancel

Discards your settings and closes the window.