Tuning Cost-Sensitive Boosting and Its Application to Melanoma Diagnosis
This paper investigates a methodology for effective model selection of cost-sensitive boosting algorithms. In many real situations, e.g. for automated medical diagnosis, it is crucial to tune the classification performance towards the sensitivity and specificity required by the user. To this purpose, for binary classification problems, we have designed a cost-sensitive variant of AdaBoost where (1) the model error function is weighted with separate costs for errors (false negative and false positives) in the two classes, and (2) the weights are updated differently for negatives and positives at each boosting step. Finally, (3) a practical search procedure allows to get into or as close as possible to the sensitivity and specificity constraints without an extensive tabulation of the ROC curve. This off-the-shelf methodology was applied for the automatic diagnosis of melanoma on a set of 152 skin lesions described by geometric and colorimetric features, out-performing, on the same data set, skilled dermatologists and a specialized automatic system based on a multiple classifier combination.
KeywordsAdaBoost Algorithm Tuning Procedure Melanoma Diagnosis Separate Cost Melanoma Data
Unable to display preview. Download preview PDF.
- 2.L. Bischof, H. Talbot, E. Breen, D. Lovell, D. Chan, G. Stone, S. Menzies, A. Gutenev, and R. Caffin, “An automated melanoma diagnosis system,” in New Approaches in Medical Image Analysis (B. Pham, M. Braun, A. Maeder, and M. Eckert, eds.), SPIE, 1999.Google Scholar
- 3.E. Blanzieri, C. Eccher, S. Forti, and A. Sboner, “Exploiting classifier combination for early melanoma diagnosis support,” in Proceedings of ECML-2000 (R. de Mantaras and E. Plaza, eds.), (Berlin), pp. 55–62, Springer-Verlag, 2000.Google Scholar
- 5.L. Breiman, “Combining predictors,” in Combining Artificial Neural Nets: Ensemble and Modular Multi-Net Systems (A. Sharkey, ed.), (London), Springer-Verlag, 1999. pages 31–50.Google Scholar
- 7.W. Fan, S. Stolfo, J. Zhang, and P. Chan, “Adacost: Misclassification cost-sensitive boosting,” in Proceedings of ICML-99, 1999.Google Scholar
- 9.J. Friedman, T. Hastie, and R. Tibshirani, “Additive logistic regression: a statistical view of boosting,” tech. rep., Stanford University, 1999.Google Scholar
- 10.C. Furlanello and S. Merler, “Boosting of tree-based classifiers for predictive risk modeling in gis,” in Multiple Classifier Systems (J. Kittler and F. Roli, eds.), vol. 1857, (Amsterdam), Springer, 2000. pages 220–229.Google Scholar
- 11.G. Karakoulas and J. Shawe-Taylor, “Optimizing classifiers for imbalanced training sets,” in Advances in Neural Information Processing Systems 11 (M. Kearns, S. Solla, and D. Cohn, eds.), MIT Press, 1999.Google Scholar
- 12.D. Margineantu and T. Dietterich, “Bootstrap methods for the cost-sensitive evaluation of classifiers,” in Proceedings of ICML-2000, pp. 583–590, Morgan Kaufmann, 2000.Google Scholar
- 14.K. Ting, “A comparative study study of cost-sensitive boosting algorithms,” in Proceedings of ICML-2000 (M. Kaufmann, ed.), pp. 983–990, 2000.Google Scholar