Quality management is the basic activity in industrial production. Assuring the authenticity of testing datasets is extremely important for the quality of products. Many visual tools or association analysis methods are used to judge the authenticity of testing data, but it could not precisely capture behavior pattern and time consuming. In this paper, we propose a complete framework to excavate the features of testing datasets and analyze the testing behavior. This framework uses min-max normalization method to pre-process datasets and optimized k-means algorithm to label the training datasets, then SVM algorithm is applied to verify the accuracy of our framework. Using this framework, we can get the features of dataset and homologous behavior model to distinguish the quality of datasets. Some experiments are carried to measure the complete framework and we use various visual formats to show these results and to verify our method.