Access trees and nodes from LightGBM model

问题

In sci-kit learn, it's possible to access the entire tree structure, that is, each node of the tree. This allows to explore the attributes used at each split of the tree and which values are used for the test

The binary tree structure has 5 nodes and has the following tree structure:
node=0 test node: go to node 1 if X[:, 3] <= 0.800000011920929 else to node 2.
    node=1 leaf node.
    node=2 test node: go to node 3 if X[:, 2] <= 4.950000047683716 else to node 4.
            node=3 leaf node.
            node=4 leaf node.

Rules used to predict sample 0:
decision id node 0 : (X_test[0, 3] (= 2.4) > 0.800000011920929)
decision id node 2 : (X_test[0, 2] (= 5.1) > 4.950000047683716)

For the Random Forest, you can obtain the same information by looping across all the decision trees

for tree in model.estimators_:
    # extract info from tree

Can the same information be extracted from a LightGBM model? That is, can you access: a) every tree and b) every node of a tree?

回答1:

Yes, this is possible with

model._Booster.dump_model()["tree_info"]

which is for example used in lightgbm.plot_tree(). I must admit though that I haven't used it myself and don't know the details about the returned structure.

回答2:

LightGBM has almost the same functions with XGBoost; sometimes I even go to the XGBoost documentation to find the functions of LightGBM. You can search for how it is done in XGBoost or you can refer directly to: https://github.com/Microsoft/LightGBM/issues/845

Also, LightGBM has a sklearn wrapper, it is probably possible to use the sklearn structure on the model you train as the way you shared. You may want to have a look at: https://lightgbm.readthedocs.io/en/latest/_modules/lightgbm/sklearn.html

Hope I could help, please do not hesitate to write if not resolved; I will go deeper in details.

来源：https://stackoverflow.com/questions/53280845/access-trees-and-nodes-from-lightgbm-model