Partition Models

Use Decision Trees to Explore and Model Your Data

The Partition platform recursively partitions data according to a relationship between the predictors and response values, creating a decision tree. The partition algorithm searches all possible splits of predictors to best predict the response. These splits (or partitions) of the data are done recursively to form a tree of decision rules. The splits continue until the desired fit is reached. The partition algorithm chooses optimum splits from a large number of possible splits, making it a powerful modeling, and data discovery tool.

Figure 4.1 Example of a Decision Tree 

Contents

Overview of the Partition Platform

Example of the Partition Platform

Launch the Partition Platform

The Partition Report

Control Buttons

Report for Categorical Responses

Report for Continuous Responses

Partition Platform Options

Show Fit Details

Specify Profit Matrix

Decision Matrix Report

Informative Missing

Actual by Predicted Plot

ROC Curve

Lift Curve

Node Options

Validation

K-Fold Crossvalidation

Additional Examples of Partitioning

Example of a Continuous Response

Example of Informative Missing

Example of Profit Matrix and Decision Matrix Report

Statistical Details for the Partition Platform

Responses and Factors

Splitting Criterion

Predicted Probabilities in Decision Tree and Bootstrap Forest