Research on Students’ Consumption Behavior Patterns

Abstract

Many studies have examined the impact of students’ learning behavior on their performance, however, consumption behaviors have not been fully explored. This study uses the consumption behavior data of 3616 students over one month to analyze. First, we processed the data and proposed the characteristics that reflect students’ living habits and campus behaviors, and conducted qualitative analysis. Second, we introduced multiple regression model to filter out features that had a significant impact on performance, and compared it with the SVR and random forests methods. The results showed that the linear regression model had a better quantitative characterization effect. We also found that some behavior patterns can distinguish students with excellent grades from those who lag behind. We further built a graph with significant features, and introduced graph convolutional neural networks for classification to verify above findings in educational data mining. Consequently, this study can effectively help tutors grasp students’ learning status and intervene promptly when necessary.