Hands-On Automated Machine Learning

Boosting

Boosting is an iterative process in which consecutive models are built one after the another based on the flaws of the predecessors. This helps to diminish the bias in the model and also leads to a decrease in variance as well. Boosting tries to generate new classifiers that are better equipped to predict the values for which the previous model's performance was low. Unlike bagging, the resampling of the training data is conditioned on the performance of the earlier classifiers. Boosting uses all data to train the individual classifiers, but instances that were misclassified by the previous classifiers are given more importance so that subsequent classifiers enhance the results. 

Gradient Boosting Machines (GBMs), which is also known ...