python-3.xscikit-learnadaboost

How to retrain AdaBoostClassifier with new data in python?


Scenario: Today I trained AdaBoostClassifier with past 1 week data and next week needs to train with new 1 week data to the existing trained classifier.

For Randomforest I'm using warm_start=True. where AdaBoostClassifier not supported directly.

https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.AdaBoostClassifier.html


Solution

  • It seems you want to perform incremental learning. In sklearn, it is not possible to perform it using AdaBoost.

    The algorithms that have implemented incremental learning are listed here. You will notice these algorithms implement the method partial_fit().

    If you want to keep using the AdaBoostClassifier, you should retrain the model using all the data (the past 1 week and the new 1 week).