Different types of classifiers combination based on choquet integral

Choquet integral, one of ensemble operators, is used widely in multiple classifiers combination when we consider the interaction between classifiers. The diversity of combination system can affect the classification accuracy and the generalization ability of the combination system. In this paper, two different types of classifiers, neural network and fuzzy decision tree, are introduced in combination system (called mix-combination system). Genetic algorithm is used to determine a non-additive set function μ which is the key issue before Choquet integral combining multiple classifiers. Some simulated experiments are run in iris, pima, cmc three datasets. The experiment results show the mix-combination systems have better performance than the combination systems only including three neural network classifiers or three fuzzy decision tree classifiers.