Abstract: This paper presents a classroom behavior management skills classification system based on machine learning to assist teachers to develop their classroom behavior management skills through guided reflection. Such a system would enable more cost-effective application of demonstrably successful approaches to having expert observers identify suggestible moments for reflection. The proposed system accepts input videos from teachers and provides classification results of specific behavior management skills that occurred on those videos. The classification results, together with relevant additional information will be provided to teachers as suggestions for reflection. The proposed approach relies on deep learning and computer vision techniques to provide the classification results. Additionally, the proposed approach has been evaluated on videos containing four of the essential teaching skills and has achieved an average F1-score of 84.75%.