The Choquet kernel for monotone data

In this paper, we introduce a kernel for monotone data de- rived from the Choquet integral with its underlying fuzzy measure. While a naive computation of this kernel has a complexity that is exponential in the number of data attributes, we propose a more ecient approach with quadratic time complexity. Kernel PCA and SVM classification are em- ployed to illustrate characteristics and benefits of the new Choquet kernel in two experiments related to decision-making and pricing.