Mainly there are two steps
- Divide the predictor space -- that is, set of possible values for X1, X2, ...Xp into J distinct and non-overlapping regions R1, R2, ...Rj
- For every observation that falls into the region Rj, make the same prediction which is simple the mean of the response values in th region Rj
- Divide the predictor space into high dimensional rectangles
- Goal is to find the regions that results in minimum mean-square-error(MSE)
- Consider the Top down greedy approach which is know as "Recursive binary splitting"
- GREEDY - In the tree building process best split is decided by considering particular step only rather than looking at how this split will affect teh further steps
- Select the predictor Xj and cutpoint s such that the predictor space results in the greatest possible reduction in the MSE
- Stopping criteria will be until no region contain some number of observations
- Check the data is pure if yes 1.a create the leaf (Either classify or predict the value)
- If data is not pure Identify the best feature to split
- Split the data
Check again whether the data is pure repeat until less minimum observations in leaf is reached