Question

Is CART the right technique to use when the dependent variable is skewed towards one of the class? If yes, can we use classification accuracy as the right performance measure? If not why? In that case, which model performance measures should be used? - Word Limit is 200 words

Answer 1

CART is Classification and Regression Trees

Yes, CART is the right technique to use when the dependent variable is skewed towards one of the class. It speaks for both classification and regression where classification is typically used when the dependent variable is binary, or with only to states and regression is for predicting values.

We are always challenged with a very rough data sets with all outliers, Skewness etc. We have to first understand the data and prepare it for analysis. If it is actually cleaning, then apply some techniques like, outlier removal, apply central limit theorem, taking multiple large samples etc. But we are using a classification technique and data is skewed towards non-responder, if we continuously split the variables into two, chances are more that same proportion might reflect till it reaches the leaf node.