The research of multi-category learning behaviors is a hot issue in interactive learning environment, and there are many challenges in data statistics and relationship modeling. We select the massive learning behaviors data of multiple periods and courses and study the decision application of regression analysis. First, based on the definition of data structure and relationship features, the data are normalized; secondly, a multi-category learning behavior prediction process is designed, the results of the data analysis are effective; computer program and Geoda are used for decision analysis, and the multiple regression equation of learning behaviors is constructed, so as to achieve the feasible prediction direction and effective decision trend. The decision application mechanism is a useful method to study multi-category learning behaviors, which can be referenced by model design and technology application for interactive learning environment.