Encouraging Cooperation in the Genetic Iterative Rule Learning Approach for Qualitative Modeling
Óscar Cordón, Antonio González, Francisco Herrera, Raúl Pérez · Studies in fuzziness and soft computing · 1999
Genetic Algorithms have proven to be a powerful tool for automating the Fuzzy Rule Base definition and, therefore, they have been widely used to design descriptive Fuzzy Rule-Based Systems for Qualitative Modeling. These kinds of genetic processes, called Genetic Fuzzy Rule-Based Systems, may be based on different genetic learning approaches, with the Michigan and Pittsburgh being the most well known ones. In this contribution, we briefly review another alternative, the Iterative Rule Learning approach, based on generating a single rule in each genetic run, and dealing with the problem of obtaining the best possible cooperation among the generated fuzzy rules. Two different ways for encouraging cooperation between rules in this genetic learning approach are presented, which are used in two different Genetic Fuzzy Rule-Based Systems based on it, SLAVE and MOGUL. Finally, the behaviour of these two processes in solving a qualitative modeling problem, the rice taste analysis, is analysed, and the results obtained are compared with two other design processes with different characteristics.