Municipal Creditworthiness Modelling by Neural Networks

The paper presents the design of municipal creditworthiness parameters. Further, the design of model for municipal creditworthiness classification is presented. The realized data pre-processing makes the suitable economic interpretation of results possible. Municipalities are assigned to clusters by unsupervised methods. The combination of Kohonen’s self-organizing feature maps and K-means algorithm is a suitable method for municipal creditworthiness modelling. The number of classes in this model is determined by indexes evaluating the quality of clustering. The model is composed of Kohonen’s self-organizing feature maps and fuzzy logic neural networks, where the output of Kohonen’s self-organizing feature maps represents the input of fuzzy logic neural networks. The suitability of the designed model is compared with other fuzzy logic based classifiers. The highest classification accuracy is achieved by the fuzzy logic neural network. The fuzzy logic neural network’s model is the best in terms of generalization ability due to this fact. By the designed model high classification accuracy, low average classification error and suitable interpretation of results are achieved..