Abstract: Using support vector machine( SVM) to classify heavy rain and hail,the classification accuracy is not satisfactory. To solve this problem,the distances of heavy rain and hail samples to the classification hyperplane are studied,as well as the neighborhood of samples and the process information of the training samples. The impact of these factors on the SVM classification credibility was analyzed. A method of defining the SVM classification credibility is put forward using distance coefficient,neighborhood coefficient and process coefficient. A classification model based on the SVM classification credibility is designed to distinguish between heavy rain and hail. Experimental results show that using distance coefficient,neighborhood coefficient and process coefficient can determine the SVM classification credibility effectively,and the Heavy rain/hail classification model is helpful to improve the POD and reduce the FAR of hail identification.