The comparison of Adaptive Neuro-Fuzzy Inference System (ANFIS) with nonlinear regression for estimation and prediction

The main purpose of the most research, especially the economic one is to access a good estimate as well as the prediction for the future. The last objective is to explore the future by which the economic plan is adopted, and the strategic policy is development. The success or failure of these plans and strategies depends on the credibility of the prediction. In spite of the Adaptive Neuro-Fuzzy Inference System (ANFIS) is characterized by being simple and flexible and have the ability to reach a perfect estimate in most cases, but the potentials in the access a good predictions are questionable. In this paper, the comparison of the (ANFIS) as an intelligence method and the Nonlinear regression (NL) as a classic method applied to define the ability of estimation and prediction. For this purpose, we choose ten nonlinear types of data, which is different in length and shape. The Anderson-Darling test is used. We conclude that six of them distributed as normal distribution while the remaining are not. By the analysis, it seems clearly that the NL is best for prediction, while the ANFIS is perfect for the estimation.