The method described in this paper enables the use of classification systems on regression tasks. The significance of this work is two-fold. First, we have managed to extend the applicability of a wide range of ML systems. Second, our methodology provides an alternative trade-off between regression accuracy and comprehensibility of the learned models. Our method also provides a better insight about the target variable by dividing its values in significant intervals, which extends our understanding of the domain.
We have presented a set of alternative discretization methods and demonstrated their validity through experimental evaluation. Moreover, we have added misclassifications costs which provide a better theoretical justification for using classification systems on regression tasks. We have used a search-based approach which is justified by our experimental results which show that the best discretization is often dependent on both the domain and the induction tool.
References
Breiman,L. , Friedman,J.H., Olshen,R.A. & Stone,C.J. (1984): Classification and Regression Trees, Wadsworth Int. Group, Belmont, California, USA, 1984.
Bhattacharyya,G., Johnson,R. (1977) : Statistical Concepts and Methods. John Wiley & Sons.
Clark, P. and Niblett, T. (1988) : The CN2 induction algorithm. In Machine Learning, 3.
Dillon,W. and Goldstein,M. (1984) : Multivariate Analysis. John Wiley & Sons, Inc.
Fayyad, U.M., and Irani, K.B. (1993) : Multi-interval Discretization of Continuous-valued Attributes for Classification Learning. In Proceedings of the 13th International Joint Conference on Artificial Intelligence (IJCAI-93). Morgan Kaufmann Publishers.
Fisher, R.A. (1936) : The use of multiple measurements in taxonomic problems. Annals of Eugenics, 7, 179-188.
Fix, E., Hodges, J.L. (1951) : Discriminatory analysis, nonparametric discrimination consistency properties. Technical Report 4, Randolph Field, TX: US Air Force, School of Aviation Medicine.
John,G.H., Kohavi,R. and Pfleger, K. (1994) : Irrelevant features and the subset selection problem. In Proceedings of the 11th IML. Morgan Kaufmann.
Kohavi, R. (1995) : Wrappers for performance enhancement and oblivious decision graphs. PhD Thesis.
Merz,C.J., Murphy,P.M. (1996) : UCI repository of machine learning databases [http://www.ics.uci.edu/MLReposiroty.html]. Irvine, CA. University of California, Department of Information and Computer Science.
Quinlan, J. R. (1993) : C4.5 : programs for machine learning. Morgan Kaufmann Publishers.
Stone, M. (1974) : Cross-validatory choice and assessment of statistical predictions. Journal of the Royal Statistical Society, B 36, 111-147.
Weiss, S. and Indurkhya, N. (1993) : Rule-base Regression. In Proceedings of the 13th International Joint Conference on Artificial Intelligence, pp. 1072-1078.
Weiss, S. and Indurkhya, N. (1995) : Rule-based Machine Learning Methods for Functional Prediction. In Journal Of Artificial Intelligence Research (JAIR), volume 3, pp.383-403.