Neural networks

A two-layer feedforward neural network (no hidden units) is applied to the same boosted dataset as was used for the decision trees. The data were transformed to ???. This network was trained using a standard learning algorihm.

Results

Learning behaviour of the neural network. As it can be seen, the Root Mean Square Error (RMSE) stabilizes in 19 learning epochs. At that moment, the network has reached the accuracy shown below.

caravan no caravan
caravan 192 51 243
no caravan 1646 2111 3757
1838 2162 4000
As it turned out, even this extremely simple neural networks learns the relevant patterns in the boosted dataset.

Results achieved with the neural net module of the Data Mining Software Kit (see: S.M. Weiss and N. Indurkhya, Predictive Data Mining, Morgan Kaufmann)