Farideh Sobhanifard, Master student industrial Engineering ,University of sistan and Baluchestan

Correct prediction of economic growth in the long-term policy planning and sustainable development, plays an important role. One of the major issues in the forecast time series is methods to identify patterns to control the complexity and Optimization of forecasting error. In this study, non-linear time series analysis GDP to forecast the economic growth path using , bayesian neural networks, for greater flexibility of the Nonlinear model in dealing with the complexities and more compatible with the actual conditions. Then, we use combination of genetic meta-heuristic algorithms to improve efficiency in network training model results and it is compared to older discussed methods. Model used 1371 to 1992 datas to estimation and tested 1393 to first two seasons of 1395 datas using the SSE and MSE. Results show that the complexity of the reform in the education network will have an important role in the deduction of optimization error.