Taken from the UCI Machine Learning Repository: http://archive.ics.uci.edu/ml/datasets/Iris

Let's build a decision tree and try it:

>>> let model = buildDTree (genMany autoDeciders) irisAttrs irisClass iris

>>> classify model [5,4,2,1]
Setosa
>>> iris !! 10
Iris {sepalLength = 5.4, sepalWidth = 3.7, petalLength = 1.5, petalWidth = 0.2, irisClass = Setosa}

Seems good! But can we really know that? Let's train and test on separate data

>>> let model' = buildDTree (genMany autoDeciders) irisAttrs irisClass (oddIx iris)

>>> dt $ confusion' model' (map (irisAttrs &&& irisClass) $ evenIx iris)
Table: Confusion Matrix
         ||-->Actual
Predicted\/             Setosa           Versicolor           Virginica
    Setosa 0.3333333333333333                                         
Versicolor                     0.30666666666666664              4.0e-2
 Virginica                    2.666666666666667e-2 0.29333333333333333

Now we see that even though not the whole data set was available when the model was induced, only few misclassifications occur.