Fuzzy support vector machine in credit rating systems

Fuzzy systems have already proven on many science areas that they can be used also for classification problems. The same situation arises in case of support vector machines. These two methods can be combined very easily although they seem to be far from each other for the first sight. In this article a new approach, namely a new Fuzzy Support Vector Machine will be introduced, which can be used for credit rating calculations. The novelty in this methods lies on the fact unlike the methods which were used before that the companies will not be separated into good or bad classes, but part of the companies belong to good and bad classes, too, but with different membership function. The target was to set up a method which gives good classification accuracy. The model is tested on data of real companies and the results are also summarized in the article. It will be proven that with use of the generating method of a specific kernel function and a specific membership function the results are very good.