Exploring the potential of skeleton and machine learning in classroom cheating detection

Abstract

In recent years, there have been certain drawbacks in the behavior of students in school cultural activities, especially the frequent occurrence of cheating in exams. Although many manual cheating detection methods have been applied, cheating behaviors have become increasingly sophisticated and difficult to detect. This makes manual cheating detection methods unable to completely prevent cheating in exams. Therefore, applying advanced technologies to automatically detect cheating cases becomes necessary to help supervisors better control the exam process. In this study, we propose a model architecture to monitor students’ cheating activities during exams. Firstly, we apply YOLO-Pose to detect skeleton keypoints, which are then passed through machine learning models such as support vector machine (SVM), decision tree (DT) classifier, random forest (RF), extreme gradient boosting (XGBoost), and propose the Ac Long short-term memory (LSTM) to detect cheating behavior. The experimental results show the feasibility of the model on various metrics. The model is tested on a dataset of cheating behaviors in the classroom, which is designed to simulate paper-based exams. Moreover, the results showed that the proposed method can detect cheating behavior in a short amount of time and can be deployed for real-world applications.