Predicting rice yield plays a significant role in preparing production plan and relevant decision-making for following year. The paper adopts ARIMA time series algorithm taking rice yield per hectare collected between 1991 and 2010 from different prefecture-level cities of Heilongjiang Province as the object of time series processing, to analyze the rice yield recorded between 1991 and 2009 of different prefecture-level cities and to build prediction model of rice yield. By using the model, the rice yield of different prefecture-level cities in 2010 is predicted. The predicted and actual space distribution of rice yield is obtained using ArcGIS software. And by conducting spatial data analysis, the space distribution diagram is mapped out. It is analyzed that the accuracy of predicted model on average attains over 95 %, featuring good prediction effect.