CODEWITHSUNDEEP
pythonscikit-learnrandom-forest
Gerges Hanna
Gerges Hanna

Reputation: 11

get the value of decision trees in random forests

Is it possible to know which number of trees get to me the predicted value? I made an object detection project using the random forests algorithm

from sklearn.ensemble import RandomForestClassifier
RF_model = RandomForestClassifier(n_estimators = 50, random_state = 42)

# Train the model on training data
RF_model.fit(X_for_RF, y_train)

Here I have 50 trees, and I need to know which number of trees get to me this predicted value.

assume random forest extract to me this is apple then i need to get the number of trees it dcided it apple like "there are 30 trees tell me this is apple and the other 20 decide it banana".

Upvotes: 0

Views: 1028

Answers (1)

swag2198
swag2198

Reputation: 2696

Yes this is possible using the estimators_ attribute:

RF_model = RandomForestClassifier(n_estimators = 50, random_state = 42)
RF_model.fit(X_for_RF, y_train)

trees = RF_model.estimators_
# Get all 50 tree predictions for the first instance
preds_for_0 = [tree.predict(X_for_RF[0].reshape(1, -1))[0] for tree in trees]

Here trees is a list of n_estimators (here 50) DecisionTreeClassifier() objects from sklearn.tree. To get the class labels from each tree, you can simply use the predict() method.

The list preds_for_0 (of length 50) stores the labels predicted for X_for_RF[0] by each constituent tree. Exploring this list will easily give you the majority label and also which tree gave what label to the instance.

Upvotes: 3

Related Questions