الفهرس | Only 14 pages are availabe for public view |
Abstract This thesis introduces a proposed Rational Fuzzy Model (RFM) to solve two basic problems in fuzzy modeling, namely rule explosion and exponential parameter growth. These two problems reduce the transparency and readability of fuzzy models. They are solved using a rational method for reducing the number of fuzzy subsets of the antecedent variables. The proposed model makes the traditional linguistic fuzzy model a special case under certain conditions. Also, the structure of the proposed RFM could be oplimi/.ed through the optimization of Us parameters. A neural implementation of the RFM (NRFM) is introduced to enable the RFM to learn from data and to extract a posteriori knowledge from these data. A tailored backpropagation learning algorithm (TBPLA) is developed to optimi/e the parameters and the structure of the NRFM. In addition, the power of genetic algorithms is utilized by devolving two genetic based learning algorithms: an adaptive genetic learning algorithm (AtjLA), and a self-adaptive genetic learning a Igorithm ( SAGLA). They a re i nlrodueed to find near global optimized parameters and structure for the NRFM. The performance of the proposed model is checked using benchmark problems. It is shown that the proposed model uses smaller numbers of input variables, input terms and output terms. Moreover, it gets the minimum mean squares error (MSE). |