Based on the multiple seasonal ARIMA model, the monthly PPI data from January 2006 to December 2018 is used to eliminate the tendency and seasonality by the phased and seasonal difference with the help of EVIEWS8.0 software. . Through comparison of indicators,. the model ARIMA(3,1,0)(1,1,1)24 is established to predict PPI from January to June in 2019．The result shows that the model has high prediction accuracy for short-term PPI, which provides the reference for the formulation of relevant economic policies.