The rule based transparent classifiers can be generated by partitioning the input space into a number of subspaces. These systems can be considered as fuzzy classifiers assigning membership functions to the partitions in each dimension. A flexible genetic algorithm based method is applied for generation of rule based classifiers. It is shown that for complex real world types of applications, a preprocessing step with neural clustering methods reduces the running time of the genetic algorithm based method drastically. A heuristic method is compared to show the strength of genetic algorithm based method.