Fracture-cavity carbonate reservoirs have strong heterogeneity and complex fluid distribution. It is often difficult to obtain parameters related to productivity prediction by conventional methods，which brings a great challenge for accurately predicting the single well productivity. This paper proposes a new method for predicting the single well productivity by selecting optimized common seismic parameters. After establishing a matrix-fracture-cave triple medium model，this paper optimizes seismic parameters affecting the single well productivity by Spearman and Pearson method，establishes the relationships between the optimized seismic parameters and the interporosity flow coefficient and storativity ratio，and introduces it into the triple medium productivity equation to predict the single well productivity of the reservoirs with different geological conditions. Based on the actual production data and preliminary test data of a fracture-cavity carbonate reservoir in Xinjiang，this paper uses the new method to predict the productivity of 134 oil wells in four areas（fault areas，underground river areas，aground river areas，and composite karst areas）and analyze errors. The results show that this method has the highest accuracy in predicting the productivity of oil wells in underground river areas，up to 87%，and for oil wells in aground river areas and composite karst areas，the prediction accuracy is low and is about 80%. The average prediction accuracy of this method is 83%，which is higher than those of multiple linear regression，BP neural network，and support vector machine. The new method makes full use of the existing seismic data，further improves the productivity prediction accuracy，and provides a new idea for predicting the productivity of complex fracture-cavity carbonate reservoirs.