Create a bootstrapped dataset

The bootstrap method is a statistical technique for estimating quantities about a population by averaging estimates from multiple small data samples.

Original dataset:

Now randomly select samples from the original dataset (This is where the name Random Forest comes from). You are allowed to pic the same sample more then once!!!

Create a decision tree

• use the bootstrapped dataset!
• only use a random subset of variables (columns) at each step!

For example we pick 2 variables (Good Blood Circ and Blocked Arteries). Lets assume Good Blood Circ does a good job in separating the samples

Now we focus on the next node:

Again, we randomly select 2 variables as candidates, instead of all 3 remaining columns: