CODEWITHSUNDEEP
pythonmachine-learningxgboost
JaneDOEE
JaneDOEE

Reputation: 11

Xgboost Underfitting

i trained the following Xgboost classifier : 

xgb.XGBClassifier(tree_method='hist', grow_policy='lossguide',gamma=1.0, max_depth=0, max_leaves=255,  min_child_weight=100,n_estimators=500,
                    n_jobs=-1,
                    learning_rate=0.1,
                    subsample=0.7,
                    colsample_bytree=0.7,
                   )

I ploted the learning curves (log loss vs epoch) for the training set and the validation set.

The learning curves that i got are exactly the same for the training and the validation set: log loss decreases at the beginning from 0.6 to 0.3 and then plateau after 100 epochs

I believe that this is a case of underfitting ? giving that i have a high bias.

How can i complexify the model in this case?

Any help is very much appreciated

Thank you

Upvotes: 1

Views: 2397

Answers (1)

Brndn
Brndn

Reputation: 846

Increase subsample and colsample_bytree.

Upvotes: 0

Related Questions