Data-based generation of fuzzy rules for classification, prediction and control with the Fuzzy-ROSA method

This paper presents three applications of the Fuzzy-ROSA method. The first application, the classification of automatic gear boxes by 149 characteristics, is an example of data-based rule generation and complexity reduction in high-dimensional search spaces. The second application is an example of the use of very noisy and contradictory data where the cancellation behaviour of insurance clients is predicted on the basis of sociodemographic characteristics. In the third application a generated fuzzy model is used to adapt the parameters of the position controller of an industrial robot to optimize the continuous path accuracy. This application demonstrates the process of learning from good and poor control strategies.