Deep learning fits a neural network for classification or regression. A random 30% of the data is used for cross-validation to find the optimal number of epochs according to cross-validation loss. The final network is trained on all data for the optimal number of epochs.
This article describes how to create a Cross-validation Deep Learning output as shown below.
- A numeric or categorical variable to be used as an Outcome variable to be predicted.
- Predictor variables that will be considered as predictors of the outcome variable.
- In the Anything menu select Advanced Analysis > Machine Learning > Deep Learning.
- In the object inspector go to the Inputs tab.
- In the Output menu select the variable to be predicted by the predictor variables.
- Select the predictor variable(s) from the Predictor(s) list.
- OPTIONAL: Select the algorithm you wish to use in the Inputs > Algorithm. By default, the Deep Learning algorithm is used.
- OPTIONAL: Select the desired Output type:
- Accuracy: When the Outcome is categorical the Accuracy outcome produces a table of accuracy by class. When the Outcome is numeric it calculates Root Mean Squared Error and R-squared.
- Prediction-Accuracy Table: Produces a table relating the observed and predicted outcome. Also known as a confusion matrix.
- Cross Validation: Produces charts of loss (e.g., network error) and accuracy or mean absolute error vs training epoch.
- Network Layers: This returns a description of the layers of the network.
- OPTIONAL: Select the desired Missing Data treatment. (See Missing Data Options).
- OPTIONAL: Select Variable names to display variable names in the output instead of labels.
- OPTIONAL: Specify the number of training rounds by inputting the desired number into the Maximum Epochs.
- OPTIONAL: Hidden layers Input a comma-delimited list of the number of units in the hidden layers.
- OPTIONAL: Select Normalize predictors for the predictor variables to be normalized to zero mean and unit variance. This is recommended if the variables differ significantly in their ranges. Note that categorical variables are also converted to dummy variables.