To explore the influence of emotion on learning is a meaningful task for educational researchers.Confusion has a great impact on the learning process,and it is of great importance to detect it effectively.In this paper,we used camera to capture the students' facial expression and defined the confusion labels through students' self-report.Then we applied a series of educational data mining methods to detect the students' confusion.Decision Tree,Na(i)ve Bayes,Support Vector Machine,Random Forest and other common classification algorithms were used to construct the model of student confusion detection.The results show that most of the used algorithm can detect students' learning confusion with at least 55%accuracy rate,and Random Forest is the best classifier with an accuracy rate of 71.18%.Furthermore,we hope to provide feedback and effective intervention to relevant students through the detection of learning confusion.