Integration of genetic algorithms and neural networks for the formation of the classifier of the hierarchical Choquet integral

Abstract

The Choquet integral model is mainly applied to describe non-additive multi-criteria decision-making (MCDM) problems. This paper considers the Choquet integral as a classifier which deals with complicated high-dimensional data. Although previously conducted studies have investigated the problem of classification using the Choquet integral and provided corresponding models, these models usually need to estimate a large number of fuzzy measure coefficients, which are not suitable when considering real situations. Sugeno et al. proposed the hierarchical Choquet integral (HCI) model to overcome this problem. However, the HCI model requires partition information of the criteria, which often cannot be obtained practically. This paper proposes two HCI models—shallow and deep models—by employing genetic algorithms (GAs) and neural networks (NNs) to automatically construct the structure of the HCI. The results of numerical experiments show that the proposed model outperforms the existing Naïve Bayes, decision tree, and NN models.