Decision Trees - Data Mining for the Masses - page 168

Figure 10-10. Predictions and their associated confidence percentages
using our decision tree.
11) We've already begun to evaluate our model's results, but what if we feel like we'd like to see
greater detail, or granularity in our model. Surely some of our other attributes are also
predictive in nature. Remember that CRISP-DM is cyclical in nature, and that in some
modeling techniques, especially those with less structured data, some back and forth trial-
and-error can reveal more interesting patterns in data. Switch back to design perspective,
click on the Decision Tree operator, and in the Parameters area, change the 'criterion'
parameter to 'gini_index', as shown in Figure 10-11.
Figure 10-11. Constructing our decision tree model using the gini_index algorithm
rather than the gain_ratio algorithm.