Investigating the emotions of high and average-performing Filipino Math learners based on electrodermal activity, facial expressions, behavior and contextual information

• Main Authors: Dy, Justin Aaron, Gorgonia, Albert Daniel, Ong, Kevin, Santiago, Sarah Kelsey
• Format: text
• Language: English
• Published: Animo Repository 2013
• Subjects: Intelligent Tutoring Systems > Philippines; Computer-assisted instruction; Computer Sciences
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/11069
• Institution: De La Salle University

Intelligent Tutoring Systems provide one way for educators to employ dier- entailed learning. This is done in order to address the various learning needs of different types of students. To be as e active as having human tutors how- ever, these systems need to be able to recognize emotions in order to respond properly to the students' learning needs. Studying and analyzing how a student behaves and feels during mathematical learning would in turn help Filipino teachers, learning-system designers, psychologists, and scientists improve the students' learning experience. The research involved the study of the active states, specifically interest, boredom, confusion and frustration of average and high performing Filipino students as they studied Math. To do this, the researchers performed data collection in different high schools to acquire facial expression, electrodermal activity, con- textual and active data. Z-score normalization was performed on the EDA data, and were segmented into 5-second windows with 3-seconds overlaps. For each of these segments, six statistical features were extracted and were synchronized with the four active labels to create the data set. Balancing was also done in order to address issues of under thing and over thing. Three machine learning algorithms, namely SVM, neural networks, and C4.5 using batch-cross validation were implemented to create the computational models. On the other hand, observations and analysis were done on the facial expressions and contextual data.