Training a fuzzy expert model with a set of data points

There are two general approaches which can be used when a fuzzy model is to be created - create model automatically on the basis of numeric data or build model manually with assistance of a domain expert. It is difficult to decide which approach gives better results because both have their own drawbacks and benefits. Sometimes, however, when expert knowledge and data knowledge are available simultaneously, both approaches can be joined together. One of the methods which is often used for dealing with this task is a method of training a fuzzy expert model with a set of data points. This method, however, is not the best alternative. The aim of this paper is to discuss drawbacks of this method and compare its performance with a performance of a method of integrating fuzzy rule nets.