Binary Classification Using Genetic Programming: Evolving Discriminant Functions with Dynamic Thresholds
Binary classification is the problem of predicting which of two classes an input vector belongs to. This problem can be solved by using genetic programming to evolve discriminant functions which have a threshold output value that distinguishes between the two classes. The standard approach is to have a static threshold value of zero that is fixed throughout the evolution process. Items with a positive function output value are identified as one class and items with a negative function output value as the other class. We investigate a different approach where an optimum threshold is dynamically determined for each candidate function during the fitness evaluation. The optimum threshold is the one that achieves the lowest misclassification cost. It has an associated class translation rule for output values either side of the threshold value. The two approaches have been compared experimentally using four different datasets. Results suggest the dynamic threshold approach consistently achieves higher performance levels than the standard approach after equal numbers of fitness calls.
KeywordsGenetic Programming Input Vector Discriminant Function Class Label Training Dataset
Unable to display preview. Download preview PDF.
- 4.Li, J., Li, X., Yao, X.: Cost-sensitive classification with genetic programming. In: Proceedings of the 2005 IEEE Congress on Evolutionary Computation, vol. 3, pp. 2114–2121 (2005)Google Scholar
- 5.Loveard, T., Ciesielski, V.: Representing classification problems in genetic programming. In: Proceedings of the Congress on Evolutionary Computation, vol. 2, pp. 1070–1077 (2001)Google Scholar
- 6.Zhang, M., Smart, W.: Multiclass object classification using genetic programming. In: Raidl, G.R., Cagnoni, S., Branke, J., Corne, D.W., Drechsler, R., Jin, Y., Johnson, C.G., Machado, P., Marchiori, E., Rothlauf, F., Smith, G.D., Squillero, G. (eds.) EvoWorkshops 2004. LNCS, vol. 3005, pp. 369–378. Springer, Heidelberg (2004)CrossRefGoogle Scholar
- 7.Frank, A., Asuncion, A.: UCI machine learning repository (2010), http://archive.ics.uci.edu/ml
- 8.Fortin, F.-A., Rainville, F.-M.D., Gardner, M.-A., Parizeau, M., Gagné, C.: DEAP: Evolutionary algorithms made easy. Journal of Machine Learning Research 13, 2171–2175 (2012)Google Scholar