Modulus genetic algorithm and its application to fuzzy system optimization

The conventional genetic algorithm encodes the searched parameters as binary strings. After applying the basic genetic operators such as reproduction, crossover and mutation, a decoding procedure is used to convert the binary strings to the original parameter space. As the result, such an encoding/decoding procedure leads to considerable numeric errors. This paper proposes a new algorithm called modulus genetic algorithm (MGA) that uses the modulus operation to resolve this problem. In the MGA, the encoding/decoding procedure is not necessary. It has the following advantages: 1) the evolution can be speeded up; 2) the numeric truncation error can be avoided; 3) the precision of solution can be increased. The proposed MGA is applied to resolve the key problem of fuzzy inference systems-rule acquisition. The fuzzy system with MGA as learning mechanism forms an "intelligent fuzzy system". Based on the proposed approach, the fuzzy rule base can be self-extracted and optimized.