Naturally-fractured carbonate reservoirs commonly exhibit complex behavior. Due to availability and a large volume of well log data, permeability prediction using these data has recently become quite popular. By finding a relationship between well logging data and FZI or permeability, the permeability value is predictable for all well depths. The relationships obtained are usually different for different wells and need to be calculated for each well separately. In this study, 169 cases of data obtained from the well logging and laboratory process related to one of the reservoirs in southern Iran have been reviewed. Using the Random Forest algorithm, the permeability of the reservoir rock with an MSE=0.009814514 has been modeled and predicted. Finally, using the automatic clustering algorithm, reservoir rock data are clustered. This process is concurrently optimized by a genetic algorithm.