Frank, E., Trigg, L., Geoffrey, H. & Witten, I. H. (1998). Naive Bayes for regression. (Working paper 98/15). Hamilton, New Zealand: University of Waikato, Department of Computer Science.
Permanent Research Commons link: http://hdl.handle.net/10289/1060
Despite its simplicity, the naïve Bayes learning scheme performs well on most classification tasks, and is often significantly more accurate than more sophisticated methods. Although the probability estimates that it produces can be inaccurate, it often assigns maximum probability to the correct class. This suggests that its good performance might be restricted to situations where the output is categorical. It is therefore interesting to see how it performs in domains where the predicted value is numeric, because in this case, predictions are more sensitive to inaccurate probability estimates. This paper shows how to apply the naïve Bayes methodology to numeric prediction (i.e. regression) tasks, and compares it to linear regression, instance-based learning, and a method that produces “model trees” - decision trees with linear regression functions at the leaves. Although we exhibit an artificial dataset for which naïve Bayes is the method of choice, on real-world datasets it is almost uniformly worse than model trees. The comparison with linear regression depends on the error measure: for one measure naïve Bayes performs similarly, for another it is worse. Compared to instance-based learning, it performs similarly with respect to both measures. These results indicate that the simplistic statistical assumption that naïve Bayes makes is indeed more restrictive for regression than for classification.
University of Waikato, Department of Computer Science.
- 1998 Working Papers