Educational Data Mining has gained a variety of attention. It describes students’ cognitive needs through data mining, and provides individualized knowledge support for cognitive differences.Although the application of data mining algorithms is relatively mature, the data pre-processing based on data collection still suffers from high costs. Thepaper focusona researchquestion: how to effectively collect behavior data in virtual learning environments?“Effectively” in the sense of ensuring that value-intensive behavior data on decision-making can be accurately collected which reflects the students’cognitive. Therefore, the paper presents a method to achieve the object.The method comprises six steps, including extraction, transformation, determination, design, trigger and store. Based on the fact that all behavior data generated by the interaction is objective, identifying the collection points on the trigger event matches the granularity level of behavior data. Considering the related platforms and intelligent applications, the method can be used, providing behavior data support for the research of knowledge services.