The discipline of data mining emerges with the development of information technology and maturation of methods of data collection,storage and management. For crop evapotranspiration (ET0 )parameter calculation has more complex and nonlinear noise characteristics,a data mining method named random forest is researched. The data mining modeling process introduced the concept called knowledge flow,through hierarchical knowledge flow model,checking knowledge flow distribution and flow situation and seizing the key link by different aspects and levels were carried out. It provided more accurate and efficient support for modeling. It will be applied to the calculation of crop evapotranspiration and verified by building a model which takes Xian weather data as input parameters and contrast it with the results from FAO56 Penman-Monteith formula. The result shows that the data mining model based on knowledge flow is of high accuracy,good stability and strong generalization ability. It can effectively predict the crop evapotranspiration. Compared to other models,it has higher efficiency and better performance,especially applicable to the prediction of large samples＇ crop evapotranspiration. It also has certain reference value for judging crop water requirements.