In this study, we propose a new variable coded hierarchical fuzzy model (VCHFM) for handling classification problems. The proposed hierarchical framework classification model synergistically integrates the standard fuzzy inference system and DNA coding with supervised learning. The VCHFM automatically generates fuzzy rules from numerical data and membership functions using both the feature extraction unit and the inference unit. Furthermore, three modified algorithms are employed by the proposed VCHFM. The implementation of this model comprises four stages. First, a genetic algorithm procedure is used to determine the distribution of fuzzy sets for each feature variable in the feature extraction unit. Second, the membership functions are adjusted by DNA computing. Third, chaotic particle swarm optimization is used to regulate the weighting grade of the principal output node in the inference unit. Finally, a multi-objective optimum fitness function is used to ensure the best classification rate with the minimum number and length of rules. We validated the proposed VCHFM by classifying five benchmark datasets: The UCI Pima Indians Diabetes, Glass, Wisconsin Breast Cancer, Wine, and Iris datasets. The computer simulation results demonstrate that the proposed VCHFM can obtain a sufficiently high classification rate, unlike other models proposed in previous studies.
All Science Journal Classification (ASJC) codes