From the course: Machine Learning and AI: Advanced Decision Trees with SPSS

Unlock the full course today

Join today to access over 23,700 courses taught by industry experts.

Pruning in QUEST

Pruning in QUEST

From the course: Machine Learning and AI: Advanced Decision Trees with SPSS

Start my 1-month free trial Buy for my team

Pruning in QUEST

“

- [Narrator] Now I'd like to discuss how Quest performs pruning. Pruning is an effort to make a smaller tree, having just grown the tree. Why would you do that? Well, it's an effort to avoid overfitting. The method that Quest uses for pruning is cost complexity pruning. If you've ever heard the phrase before, perhaps you're familiar with CART. CART uses the same method. Another name for this is weakest link pruning, which I really think captures the idea. The idea is to identify branches in the tree that don't help increase the accuracy of the tree. We're doing a bit of housekeeping, so to speak. Let's take a look at it. In interactive mode, you can actually grow the tree without pruning, and then apply the pruning. Let's take a look. Without the pruning, the tree is actually quite large in many cases. So you can see here that it's grown one, two, three, four, five levels. We can zoom in a little bit and take a look. It split first on passenger class, not surprising given what we know…

Contents