Combined Forecasting Model Based on Self-Adaptive Filtering Weights

The appropriate combination of different forecasting models tends to be more accurate than the forecasting of a single model. Based on the self-adaptive filtering method, a method of determining the "best" weights of each single model by self-adaptive iterative adjustment of weights is proposed. Then, the combined forecasting model based on self-adaptive filtering weights is established by weighting each single forecasting model. The GM(1,1), BP neural network, trend extrapolation method forecasting models, error absolute value reciprocal method and self-adaptive filtering weights method combined forecasting models are used to forecast China's GDP data. The results show that the combined forecasting model based on self-adaptive filtering weights has higher forecasting accuracy than other models and forecasts GDP from 2018 to 2021.