Problem-Based Learning in Data Science Course: Analysis of Online Learning during the Covid-19 Pandemic

This paper presents the Live Interactive Streaming Classroom (LISC) system model which can be applied to the Problem-Based Learning (PBL) and motivate students to solve problems in a real-life setting with guidance from teachers via online learning. In this study, the students used self-learning methods for interactive class. The class was managed through the online streaming system by the video conference applications. The study showed that the students' coding and real-life problem solving skills were improved from hand-on experience. The LISC system model stimulated students' learning levels. This was proved by applying the proposed method to the Data Science class at Silpakorn University, Thailand, joined by an instructor from Japan. When the course ended, the evaluation from the participating students showed that they were much satisfied that their skills were up-leveled for coding and real-life problem solving, and their attitudes towards the Data Science class improved. It implied and confirmed that the proposed method was efficient to the students' practical learning level, among the new normal living period in the Coronavirus pandemic. © 2022 ACM.

Emotion recognition of students during e-learning through online conference meeting

Due to the outbreak of COVID-19, online learning has become a way of life. The objective of this study was to propose techniques to detect students' emotions while studying via online video conferencing. This proposed technique, which updates the facial emotion image of the current class-member, enables the system to achieve a highly accurate performance for facial emotion recognition. This proposed technique can be applied to online teaching systems. As a result, instructors can identify the interest levels of each learner using the interest assessment system, which measures and monitors the period of tiresomeness of each learner. The results showed that our techniques achieved a high percentage of accuracy for each emotion, that is, sleepy/bored = 93.3%, confused = 94.3%, neutral = 92.6%, and happy = 97.2%, which was higher than the convolutional neural network-based emotion recognition system. The proposed system was applied to a real class and satisfactory overall results of 88.7% were achieved. This study proved the feasibility of the proposed technique. © 2021 Silpakorn University. All right reserved.