Why height of tree is one?

[Piyushbalwani]

Hi,

If height of HoeffdingTreeClassifier is 1, what does it implies? Is data not sufficient to converge or data is not good?
This problem is reproducible for many small datasets. The accuracy and F1-score is good for the same HoeffdingTreeClassifier model, but still None dataframe returned by the model.
Data is sufficient for the sklearn DecisionTreeClassifier and it seems to be fit nicely and have decent height for prediction.

Note: accuracy and f1 score of HoeffdingTreeClassifier is calculated during sample wise learning of the classifier.

Thanks!!

[jacobmontiel]

Without information about the data (size, type, etc.) it is hard to provide a precise explanation. In the example that you describe, it is likely that the tree has not "seen" enough information to make a split. The batch Decision Tree classifier, on the other hand, may grow as it has access to all the data. The Hoeffding Tree model approaches the performance of the batch Decision Tree model as the number of samples increases.

In other words, the Hoefdding Tree grows as it gets more data, and the split criterium is met (Hoeffding bound).