AdaBoost

CS 301 ? Intro to Data Mining Ensemble Methods (continued) ? Use multiple classifier models (built using a single method like ID3) to obtain better predictive performance than could be obtained from any of the individual classifier models ? Methods for Constructing an Ensemble Classifier (1) Specify classifier method you want to use (e.g., ID3, PRISM, etc.) (2) Manipulate the training dataset according to some strategy - multiple training sets created by resampling the original data according to some sampling distribution (3) Build a base classifier from each training set (using the classifier method you specified) (4) Construct an ensemble classifier by considering how the base classifiers would make predictions on original dataset (consensus opinion)