Abstarct: Discrete Fracture Network (DFN) model is a random display of natural fractures network. Fracture network is produced by conducted statistical measured data. In DFN model, a unique collection of fractures baxsed on random descxriptions using fractures properties such as fracture intensity, orientation and size are produced. This study has been done in two stages on seven wells of a carbonate hydrocarbon reservoir located in Southwest of Iran. In the first step, the fractures of the area and its fracture network model built by DFN using image log fractures data. But the lack of data and image log fractures related to the complex nature of fractures causes, resulting models are always associated with uncertainty and smoothing. In the second step, using gamma-ray logs data, the systemtracts of sequences in wells is determined. the study analyzed data from wells with fracture and evaluate the severity of fractures and the ratio of fracture zone to not broken zone of systemtracts of wells by artificial neural network, new fracture intensity graphs of a wells were built. Then, using these graphs, the new DFN was made. In fact, for DFN models which had been made in the previous step, the sequence is determined by condition. Improved modelling to calculate the correlation coefficient was used as a measurement to quantify the results and compare two methods in each well. Comparing correlation coefficients for first and second steps models was illustrated that the correlation coefficient for second model in wells B, C, D and E improved respectively 0.16, 0.47, 0.16 and 0.14. So first step DFN is improved.