Introduction

This demo version, after briefly presenting the regression trees, focuses on pruning. Pruning a regression tree (which consists in eliminating some well-chosen parts of the tree) is essential in order to get an optimal tree. Pruning therefore permits to avoid two potential issues:

• Reaching a too large tree leading to overfitting; that is obtaining a too complex model that fits the random noise included in the observations instead of the underlying pattern of these observations
• Having a too small tree leading to high bias

In order to optimize the predictive power of the model, the optimal balance between variance (high with a too large tree) and bias (high with a too small tree) has to be found. This is the final goal of the pruning technique.

Author: Julien Antunes Mendes  | Publisher : Reacfin Academy

Demo 

Click on next to begin the demo