Week 1 -- Decision Trees

We covered this in class, and we’re going to cover it in detail during our first Saturday review session, so I’m only going to skim over it here. The most important thing you need to know:

The point of a decision tree is to help us make a decision.

Sounds obvious, right? But think about what that means. It means that while we care about the odds of getting hit by a car while walking to work, we care more about using these odds for something useful, namely, deciding whether it might be safer to drive, take the tram, etc. A decision tree, then, is a tool that classifies our choices in a way that helps us make a decision. That’s all it is.

Decision trees always start with a choice, i.e. a square node, also called a decision node. After all, the point of all this is to tell us which choice(s) to make. Circle nodes (defined in this class as nature nodes) give us nature’s chance selection of a particular outcome occurring.

I would use different words to describe ‘nature' nodes. When I think of nature, I think of trees and parks… especially since I associate these with decision ‘trees,’ which are in fact not trees at all. So I think it might be confusing. If I could, would call these nodes ‘random’ nodes, or something like that. As in, if you were trying to figure out whether to gamble, this node would tell you that a coin will randomly come up heads 50% of the time. If we put money in the stock market, we will gain money 60% of the time, that sort of thing. Each circle node contains a probability, which is random because we have no control over it… even though we have the choice as to whether we want to pursue the path.

Don’t worry about the terminology too much. In terms of designing a tree, take a look at the practice problems, and we’ll spend significant time on this during the first review session on February 21st. Bring your questions then, because they’re hard to answer in this forum.

^_^