Learning Decision Trees Adaptively from Data Streams with Time Drift

We propose a new method for mining concept-drifting data streams using decision trees and adaptive windowing. We present a new algorithm based on Hulten-SpencerDomingos’s CVFDT that overcomes some of the shortcomings of CVFDT, specifically, dependence on user-entered parameters that determine the guessed speed of change. Our algorithm detects when change occurs and provably adapts to the speed of change without user intervention. It is based on ADWIN, an adaptive algorithm for detecting change and maintaining an updated sample from the input sequence automatically. Our experiments show that the new algorithm does never worse, and in some cases much better, than CVFDT.