Neural network (MLP)
A small multi-layer perceptron for classification or regression, with a holdout fit.
When to use it
Use the MLP when you want a small neural net on tabular data, as classification or regression. Trees with readable importances are gradient boosting or random forest.
Assumptions
Hidden-layer size is not a user control. The fit uses two hidden layers of 64 and 32 units. This run also prints permutation importances. Holdout fraction 0.25; HTTP and chat dispatch both honour test_fraction.
Running it in Tensr
Options
Prop
Type
Reading the output
The same strong pass/fail sample, classification mode, hidden layers 64 and 32, 25% holdout.
| Actual \ Predicted | 0 | 1 |
|---|---|---|
| 0 | 25 | 8 |
| 1 | 11 | 16 |
| Variable | Importance |
|---|---|
| hours | 0.193 |
| anxiety | -0.018 |
Train accuracy = 0.856, test accuracy = 0.683, n = 240. Hours was built to move the pass rate strongly, so holdout accuracy should sit clearly above chance.
Reporting (APA 7)
Holdout classification of passed from hours and anxiety, n = 240. Test accuracy = 0.683, train accuracy = 0.856.
Coming from SPSS
SPSS Neural Networks is Analyze → Neural Networks. Tensr’s items are on the ML menu. The path string stored for this item says ML → Neural Network. Hidden layers are not a dialog control.
Related
Gradient boosting is the tree ensemble with the same mode switch.