SANN - Results - Details Tab

Neural Networks

Select the Details tab of the SANN - Results dialog box to access the options described here. For information about the options that are common to all tabs, see SANN - Results.

Summary. Click the Summary button to generate a spreadsheet containing the summary details listed in the Active neural networks grid box.

Weights. Click the Weights button to display a spreadsheet of weights and threshold for each model in the Active neural networks grid.

Correlation coefficients. Click the Correlation coefficients button to generate a spreadsheet of correlations coefficients for the target variables (for train, test, and validation samples) using the active neural networks models. Note that this option is only available for regression and time series (regression) analyses.

Confusion matrix. Click the Confusion matrix button to generate a confusion matrix and classification summary for the categorical target. A confusion matrix gives a detailed breakdown of misclassifications. The observed class is displayed at the top of the matrix, and the predicted class down the side; each cell contains a number showing how many cases that were actually of the given observed class were assigned by the model to the given predicted class. In a perfectly performing model, all the cases are counted in the leading diagonal. A classification summary gives the total number of observations in each class of the target, the number of correct and incorrect predictions for each class, and the percentage of correct and incorrect predictions for each class. This information is provided for each active network. Only the samples selected in the Sample group box (e.g., Train) will be used in generating the two spreadsheets. Note this option is only available for classification and time series (classification).

Confidence levels. Click the Confidence levels button to display a spreadsheet of confidence levels for each case in the sample selected in the Sample group box. Confidence levels will be displayed for each model. You can specify which cases to include in the spreadsheet using the options in the Sample group box. Note that this option is only available for classification problems.

Predictions statistics. Click the Predictions statistics button to generate a spreadsheet containing minimum and maximum prediction values, residuals, and standardized residuals for each model in the Active neural networks grid. These statistics will be reported for each sample (training, testing and validation). This option is only available for regression and time series (regression) analyses.

Data statistics. Click the Data statistics button to generate a spreadsheet containing the mean, standard deviation, minimum value and maximum value for each continuous variable in the analysis. These data statistics will be broken down by each sample (training, testing, and validation) and also reported for the overall data set.

Global sensitivity analysis. Click the Global sensitivity analysis button to conduct a sensitivity analysis on each model and display the results in a spreadsheet. Sensitivity analysis rates the importance of the models' input variables. You can conduct sensitivity analysis on a per sample basis, using the options in the Sample group box.

Note:

Global sensitivity for continuous input: Error when input is set to mean divided by error when input is used.

Global sensitivity for categorical input: Average error when input is set to all other categorical levels divided by error when input is used.

For regression, error is sum of squares. For classification, error is cross-entropy. If variable is important, global sensitivity should be large (>>1).

Local sensitivity analysis. Click Local sensitivity analysis button to generate a separate spreadsheet of local (pointwise) sensitivity analysis for each model in the Active neural networks grid. Local sensitivity analysis indicates how sensitive the output of a neural network is to a given domain of an input variable. These sensitivity values are actual first-order derivatives evaluated at specific centile points for each input. For each input the derivative is taken with respect to the target at ten evenly spaced locations with the observed minimum and maximum values serving as end points. Other input variables are set to their means during this calculation. A separate spreadsheet is also generated for each dependent (target) variable as well. You can conduct pointwise sensitivity analysis on a per sample basis, using the options in the Sample group box. Note that this option is only available for regression analyses with no categorical inputs.