A GA-BP neural network for nonlinear time-series forecasting and its application in cigarette sales forecast

Abstract Neural network modeling for nonlinear time series predicts modeling speed and computational complexity. An improved method for dynamic modeling and prediction of neural networks is proposed. Simulations of the nonlinear time series are performed, and the idea and theory of optimizing the initial weights and threshold of the GA algorithm are discussed in detail. It has been proved that the use of GA-BP neural network in cigarette sales forecast is 80% higher than before, and this method has higher accuracy and accuracy than the gray system method.