Abstract: Accurately predicting the prices of agricultural products is very important for evading market risk，increasing agricultural income and government macroeconomic regulation. With national cotton prices prediction asan example，this paper proposed a SVM prediction model of agricultural products prices based on fuzzy informationgranulation and particle swarm optimization algorithm. Firstly，the original national cotton prices A index time seriesdata were transformed into fuzzy information granulation particles made up of Low，R and Up. Secondly，particle swarmoptimization algorithm was used to find the best parameters c and g for SVM model. Finally，cotton price fluctuation rangeand change trend in the future were predicted by the optimized SVM regression model. The empirical analysis showedthat the PSO-SVM model was effective for the prediction of cotton price fluctuation range and change trend.