How decision tree handles the missing value

How does the decision tree handles the missing value

Hello @vivek16pawar,

First, you have a specific decision-tree-based method package that you are interested in (e.g. xgboost, lightgbm, catboost, …), then google for example “xgboost handle missing values”, and then look for the package’s official documentation about how it handles missing values.

For example, this is the google result for xgboost which essentially says that a feature’s missing values are left as is (won’t be imputed), and whether to put those samples (with missing value in that feature) to the left or the right child node will be based on the splitting criteria.