A support vector machine prediction method based on the surveillance of case-based reasoning is presented aimed at the problems of small sample, small range and data distortion during the military aircraft NPD data integration. This method adopts support vector machine as regression and prediction model, searches similar case via case-based reasoning and uses it to supervise the process of regression and prediction. In this process, a method of comparability search, checkout and revise based on gradient is put forward successfully to combine case-based reasoning and prediction model, which reduces the sensitivity and dependence of model on the data sample. An example indicates that the mothed is economical and effective, which will greatly improve the precision of regression and prediction.