| Building trees interactively | Integration by the Runtime Engine (Windows) | |
When using a discrete outcome the resulting tree is referred to as a classification tree. Associated with each leaf of the classification tree is an outcome and the probability of that outcome being the outcome of a data record falling into that leaf profile. Predictive rules are represented by leafs with high probability figures.
By contrast, when using a continuous (numeric or date/time) outcome the resulting tree is called a regression tree. Associated with each leaf of a regression tree is an average value for a data record falling into that leaf and the standard deviation from that value. Predictive rules are represented by leafs with low standard deviation figures. Applying a regression tree to a data record will generate a prediction of the average value of its outcome and an estimate of the margin around the average value that the outcome value might fall.