A new forecasting approach with neuro-fuzzy architecture

Planning is an integral part of any work process. However, it is difficult to plan effectively if uncertainties cloud the planning horizon. Forecasts can help organizations by reducing some of the uncertainty, thereby enabling them to develop a more practical plan. Currently, there are many different kinds of forecasting techniques available, but no single technique works best in every situation. Therefore, the objective of this paper is to propose a new intelligent forecasting technique which employs the concepts of neural network and fuzzy system. The neural network determines the output of the system based on the current state of the input parameters. The fuzzy-inference network evaluates the performance of the model by assessing the error and the derivative error of the system. If the error is high, the corrective action will be sent to the neural network to improve the system performance. To test the performance of the proposed model, it is used to approximate the nonlinear function in comparison to the most commonly used backpropagation neural network. The testing results demonstrate a very reliable performance of the proposed model.