Neuro-fuzzy extraction of interpretable fuzzy rules from data

The paper addresses extraction of linguistic fuzzy rules from data, paying specific attention to such properties of the resulting fuzzy model as interpretability and generalization ability. A modeling technique, combining some previously known heuristic modeling approaches, is developed. Experiments of controller identification based on the truck backer-upper application demonstrate that the proposed technique is able to capture the relevant information even if the data sets used for model extraction are insufficient and/or contain noise.