Classification

The classification table shows the practical results of using the network. For each case, the predicted response is the category with the highest predicted pseudo-probability.

• Cells on the diagonal are correct predictions.
• Cells off the diagonal are incorrect predictions.

Given the observed data, the "null" model (that is, one without predictors) would classify all customers into the modal group, Plus service. Thus, the null model would be correct 281/1000 = 28.1% of the time. The RBF network gets 10.1% more, or 38.2% of the customers. In particular, your model excels at identifying Plus service and Total service customers. However, it does an exceptionally poor job of classifying E-service customers. You may need to find another predictor in order to separate these customers; alternatively, given that these customers are most often misclassified as Plus service and Total service customers, the company could simply try to upsell potential customers who would normally fall into the E-service category.

Classifications based on the cases used to create the model tend to be too "optimistic" in the sense that their classification rate is inflated. The holdout sample helps to validate the model; here, 40.2% of these cases were correctly classified by the model. Although the holdout sample is rather small, this suggests that your model is in fact correct about two out of five times.

Next