I am using zip codes as an independent variable as part of a binary classification problem. Naturally, this feature has many different levels (around 2,000), so I was wondering if there is a standardized method for reducing the number of categories by grouping them together using a decision tree.
The reason I specifically want to tackle this problem using a decision tree is didactical. I know of other methods, but have never used a decision tree for this purpose.
To be more specific, I would like to reduce the number of levels from 2,000 down to 20 or so different groups.