Abstract:Many factors may cause the fluctuation of agro-product market prices,so the agro-product price often experience ups and downs,whose fluctuations are similar to the random walk.In order to scientifically analyze and predict the trend of daily price of agro-product market,this paper selected the daily wholesale price of tomatoes in China as object to model,and the data used in the modeling are between 2008 and 2009 with daily prices for 731 days.The ultimate goal is to provide technical support for price forecasting and market regulation.According to the random features of daily price fluctuation of agro-product market as well as ADF test and ARCH effect test based on price series data,this paper employed the modern time series methods of ARIMA,ARCH and GARCH to establish daily wholesale price forecasting models of tomatoes respectively,and applied the models to forecast the tomato price from January 1,2010 to January 10,2010 as evaluation.The result shows that mean absolute percentage error (MAPE) of the three daily price forecasting models is less than 2%,among which the highest accuracy in forecasting is GARCH model.Accuracy of the three models forecasting is ideal if unexpected incidents don’t occur in super short-term agro-product market price forecasting.But it is hard to simulate and forecast quantitatively for emergencies causing dramatic changes.