Classroom management has many challenges, but when the class is held in a computer lab, even more challenges surface. The obstruction of the line-of-sight between the instructor and the students makes it difficult to monitor student behavior, attention, facial expressions or other non-verbal cues. Also, while the computer is a powerful machine, and can enhance the learning experience if used correctly, access to games and a web-browser makes it a potential distraction. Advancements in areas such as the Internet-of-Things (IoT), context-aware systems (C-AS), and machine learning (ML) can and have been leveraged to provide a solution. The rise of smart classrooms, which can utilize any subset of IoT, C-AS and ML, has aided in classroom management in many different ways. Many implementations have specific or limited functionality. Others only operate under a narrow context, or a single learning activity. While these systems are extremely useful or powerful in their context, they do not address the fact that many different types of learning activities can take place in the classroom. However, the system proposed in this work attempts to address some of these issues in classroom management, and to operate under multiple learning activities. The proposed system is a C-AS, whose context model was inspired by the phases of the Gradual Release of Responsibility framework (GRR) for learning, monitors student behavior and interaction with the computer in a computer lab setting in order to determine his/her level of attentiveness. This study followed a proposed pervasive computing system design process, which was based on the more mature embedded systems design process. The system uses input from the instructor to determine which sensors to use, how to process the data from those sensors, and which trained ML models to use. The proposed system was tested for two learning activities during run-time, a lecture and a computer-based exam, using computers with a standard RGB webcam, and one with a RealSense camera array. The overall best performing model was the Random Forest classifier fit to the RealSense camera data for the lecture, with around 94% accuracy. The proposed system shows promise as a tool to enhance classroom management under multiple learning activities. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com.bibliotheek.ehb.be/en-US/products/dissertations/individuals.shtml.]