To cope with concept drift, we placed a probability distribution over the location of the most-recent drift point. We used Bayesian model comparison to update this distribution from the predictions of models trained on blocks of consecutive observations and pruned potential drift points with low probability. We compare our approach to a non-probabilistic method for drift and a probabilistic method for change-point detection. In our experiments, our approach generally yielded improved accuracy and/or speed over these other methods.
Stephen H Bach (University of Maryland)
Mark Maloof (Georgetown University)
More from the Same Authors
2012 Poster: Scaling MPE Inference for Constrained Continuous Markov Random Fields with Consensus Optimization »
Stephen H Bach · Matthias Broecheler · Lise Getoor · Dianne P O'Leary