Abstract:In order to study the changing trend of China’s residents consumption level to predict the residents consumption level in next four years (2020—2023), this paper selects the samples of China’s residents consumption level in 18 years from 2002 to 2019, constructs induced ordered weighted arithmetic averaging (IOWA) operator combination forecast model based on ARIMA (2,1,1) model, HoltWinters no season model and multivariate regression model, mainly uses the combined prediction optimization method with minimizing error square sum to predict the residents consumption level in next four years (2020—2023), evaluates the effectiveness of induced ordered weighted arithmetic averaging (IOWA) operator combination prediction model, searches explanatory variables and sets up multivariate regression model based on inner cycle and outer cycle of new development pattern. Research finds that the prediction accuracy of IOWA operator combination prediction model is higher than that of single prediction model and each error is significantly lower than that of the single prediction model. In the next four years, the consumption level of China’s residents will not fluctuate largely but tend to decline in a small range.