Combination Model Research Based On Dynamic Regression Model

Time series analysis is a branch of the application of probability and mathematical statistics, which follows the basic principle of mathematical statistics. It can be adjusted according to the new data obtained, so as to determine the parameters of the model, to improve the prediction accuracy of the model. However, because of the impact of various factors, they can only be used to analyze and predict the overall trend of the data, and can’t be fully analyzed.In this paper, by the study of combination model based on the ARIMA model and the dynamic regression model, we combined the number of inbound tourism in Zhejiang Province from 1994 to 2014 and date from Combination model research based on ARIMA, and studied the further data. We find that the combination has higher prediction efficiency.First, we study the correlation theory of the time series model and the related knowledge, and analyze the ARIMA model and the dynamic regression model. Whether the data is able to establish the model, and the rationality of the model is given.Secondly, the paper establishes the ARIMA model and the dynamic regression model to verify the rationality of the model, and predicts the data. The forecast results are analyzed based on three evaluations:relative percentage error, mean absolute percentage error, mean absolute error. The dynamic regression model has higher efficiency of prediction than exponential smoothing method.Finally, we choose four methods of the combination model, and the selection of the model. The selection method of the weight coefficient bases 5 kinds of evaluation index, which are SSE, MSE, MAE, MAPE and MSPE. The combination model has breakthrough success in the model selection. However, during compare four combination model to each other, we need further research to select the best model.