An Immuno-Genetic Hybrid Algorithm

The construction of artificial systems by drawing inspiration from natural systems is not a new idea. The Artificial Neural Network (ANN) and Genetic Algorithms (GAs) are good examples of successful applications of the biological metaphor to the solution of computational problems. The study of artificial immune systems is a relatively new field that tries to exploit the mechanisms of the natural immune system (NIS) in order to develop problemsolving techniques. In this research, we have combined the artificial immune system with the genetic algorithms in one hybrid algorithm. We proposed a modification to the clonal selection algorithm, which is inspired from the clonal selection principle and affinity maturation of the human immune responses, by hybridizing it with the crossover operator, which is imported from GAs to increase the exploration of the search space. We also introduced the adaptability of the mutation rates by applying a degrading function so that the mutation rates decrease with time where the affinity of the population increases, the hybrid algorithm used for evolving a fuzzy rule system to solve the wellknown Wisconsin Breast Cancer Diagnosis problem (WBCD). Our evolved system exhibits two important characteristics; first, it attains high classification performance, with the possibility of attributing a confidence measure to the output diagnosis; second, the system has a simple fuzzy rule system; therefore, it is human interpretable. The hybrid algorithm overcomes both the GAs and the AIS, so that it reached the classification ratio 97.36, by only one rule, in the earlier generations than the two other algorithms. The learning and memory acquisition of our algorithm was verified through its application to a binary character recognition problem. The hybrid algorithm overcomes also GAs and AIS and reached the convergence point before them.