STUDY ON MONTHLY CENTRAL TAX REVENUES FORECASTING MODELS BASED ON TIME SERIES METHOD
基于时间序列法的国税月度收入预测模型研究

The monthly central tax revenues forecasting is considered based on time series method. A monthly tax revenues forecasting method, including how to fit, test, forecast, evaluate and dynamically revise the model, is proposed through the combination of ARIMA model proposed by Box-Jenkins and central tax data. According to this method, monthly tax revenues of value-added tax, customs duty and business tax from January to December in 2006 are predicted. Comparing with actual monthly revenues of these three taxes during 2006, the average relative errors are no more than 5.47\%, 8.63\% and 2.37\%, respectively. Finally, suggestions of dynamically revising the forecasting model in practical application are presented. Problems that occur during the use of time series method in tax revenues forecasting are also simply discussed.