One of the important issues in the economy is the prediction of economic growth. The correct prediction of economic growth has an important role in government policy and economic planning and can help policymakers in future decision making. In this research, nonlinear structure with non-linear data on the economic value added of non-linear models based on adaptive-fuzzy networks (ANFIS) is used to predict the GDP of Iran. To this end, seasonal data on GDP in Iran have been used from spring 1385 to first quarter of 1395 to predict GDP without oil and its growth, divided into three parts: industry, agriculture and services. Neural-fuzzy network modeling has been investigated in three models with different membership functions. Finally, by prioritizing the power of models in prediction, in each of the three sections, the best model is derived according to the error function. The results of the test data for the first nine months of the year of 1395 show the accuracy of the ANFIS model in the prediction. the error values of each type, ANFIS model to predict future values of GDP of non-oil sectors of industry, agriculture and services and for the nine courses forward this index has been predicted. According to the obtained values, the country's non-oil economic growth in 1395 to about 6 percent. this amount in the years 1396 and 1397 respectively 2 and 3. 5 percent.