Educational Data Mining toward Personalized Tutoring: Exploration of Facial Behaviors, Thermal Comfort, and Relevant Content Search

Jiang, Han

Etd

Educational Data Mining toward Personalized Tutoring: Exploration of Facial Behaviors, Thermal Comfort, and Relevant Content Search

Public Deposited

We present work on machine learning and educational data mining with the long-term goal of helping to personalize students’ learning experiences. The first part is using machine learning methods to analyze videos of students in educational settings. In one project, (1) we explored the relationship between students' thermal comfort, engagement, and learning in a laboratory experiment video dataset and built an end-to-end detector to measure students’ thermal comfort and engagement from their faces. In another, (2) we investigated if the empathic messages provided by an intelligent tutoring system could influence the students’ emotions and heart rate. In a third, (3) we built a model for predicting when human teachers shift their eye-gaze to look at their students during 1-on-1 math tutoring sessions. The second part is about personalizing educational content by analyzing the detection results of educational videos on YouTube: (4) we compared different methods to provide better math tutorial video recommendations to students by ranking the videos based on the representations conducted by detected math information or the transcripts. Along the way, (5) we found a new kind of training set bias based on the mathematical correctness of object configurations in a visual scene, particularly within the context of detecting individual symbols in images from math tutorial videos.

Creator