Workspace Node: SANN Regression/Classification/Time Series - Specifications - Subsampling Tab

In the SANN Regression, SANN Classification, SANN Time Series (Regression), or SANN Time Series (Classification) node dialog box, under the Specifications heading, select the Subsampling tab to access the following options. This tab is available only if Subsampling (random, bootstrap) is selected on the Specifications - Quick tab. See also, SANN Overviews - Ensembles and Subsampling.

Element Name Description
Subsampling method Use the options in this group box to specify which subsampling method you want to use.
Random Select this option button to generate multiple subsamples of the original data set based upon sampling without replacement.
Bootstrap Select this option button to generate multiple subsamples of the original data set based upon sampling with replacement.
Size of subsamples (%) Use the options in this group box to specify the percentages of cases in the data set to allocate to the train, test, and validation samples.
Train subsample Specify the percent of valid cases to use in the training sample. Must be greater than 0 and less than or equal to 100. This option is not available for deployment.
Test subsample This option is to randomly assign cases to a test sample. Specify here the percentage of cases to use. Zero sample size means no test sample is requested (not recommended).
Validation subsample This option is to randomly assign cases to a validation sample. Specify here the percentage of cases to use. This option is not available for deployment. Zero sample size means no test sample is requested (not recommended).
Number of subsamples Specify the number of subsamples you want to create. This is equal to the number of neural networks that will be created, that is, one neural network is created for each subsample. In theory the more subsamples (and, hence, neural networks) you create, the better, but from a practical standpoint, you should limit this number depending on your time and computer resources.
Seed for subsampling The positive integer value entered in the Seed for subsampling box is used as the seed for a random number generator that produces the random subsamples from the data. Starting from the same seed will yield the same subsample. If you want to create a different data subsample, change the seed value.

Options / C / W. See Common Options.

OK Click the OK button to accept all the specifications made in the dialog box and to close it. The analysis results will be placed in the Reporting Documents node after running (updating) the project.