Online learning and testing formats have developed rapidly in developed countries, allowing students to control their learning time. This paper introduces a facial analysis system to predict student concentration levels during online learning and integrates the results into a learning management system (LMS). The system is built based on facial recognition using CNN networks, detecting 68 facial landmarks and applying rules to track gaze direction to monitor students’ concentration during online learning. Student concentration and study frequency data will be used to analyze the correlation with learning outcomes. Experiments conducted on the Posts and Telecommunications Institute of Technology (PTIT) dataset with 1674 images and gaze data from MPIIGaze achieved an accuracy of 96.21% in face recognition and 84.27% in gaze tracking, with a processing speed of 46 ms. Experimental results from 99 students show a positive correlation between concentration levels, study frequency, and students’ performance. The system demonstrates the potential for integration into LMS systems to improve monitoring and enhance the effectiveness of online learning.