Abstract: Water resources is one of the important factors impacting food production. It's a hot issue to investigate the response of grain yield to climate change driving water resources change. Taking winter wheat and summer maize in Fenhe River basin as the study object, six machine learning algorithms were used to build the yield prediction models, including Linear Regression, Back Propagation Neural Networks, Support Vector Machine Regression, Random Forest, Radial Basis Function, and Extreme Learning Machine. Based on the response of water resources to climate change by the climate-elasticity coefficient, the comprehensive response of grain yield was analyzed on a catchment scale, which was related to climate change driving water resources change. The results indicated that ① Machine learning algorithms performed well on the simulation yield of the winter wheat and summer maize in Fenhe River basin. ② There was a 19.4% increase in water resources as a 10% increase in precipitation, otherwise, a 1℃ increase in temperature might lead to a 4.3% decrease in water resources. ③ If precipitation decreased by 10%-30%, the yields of winter wheat and summer maize would decrease by 6.4%-19.3% and 4.0%-15.0%, respectively. ④ When the temperature increased by 0.5-3.0℃, there might be 1.8%-17.1% and 1.2%-7.9% increases in the yields of winter wheat and summer maize, respectively. ⑤ The yield of winter wheat was more sensitive to climate change than that of summer maize in the Fenhe River basin. The results were useful for future adaptive strategies of water resources management and agricultural production.