鄭偉凱

Emotional problems of students with autism play an important role in their learning in mathematics e-learning environments. This study proposed an emotional adjustment model for students with high-functioning autism in mathematics e-learning with an emotion classification mechanism. The present paper reports the development of the emotion classification mechanism through evoking autistic students’ emotions in a mathematical e-learning environment and recording changes in their physiological signals and facial expressions. A total of fifty-eight measures were obtained from an experiment, and twenty-nine measures were further extracted from one-way ANOVA and information gain (IG) methodology. Support vector machines (SVM), k-nearest neighbors (KNN), and classification and regression trees (CART) were combined with bootstrap aggregating (Bagging) individually and used to classify four emotional categories: calm, happy, anxious, and angry from the twenty-nine features. The accuracy rate of the SVM ensembles using Bagging to classify emotions reached the highest (81%). The emotion classification mechanism developed in the present report could support the emotional adjustment model which aims at classifying and adjusting autistic students’ emotions during mathematics learning and enhancing their learning effectiveness.