Assessing Students’ Emotions in Learning Environment: Integrating Deep Learning with Statistical Learning

Understanding the facial expressions is one of the most widely utilized method to discover the hidden emotions of an individual. Ascertaining the emotions are helpful across most of the domains ranging from psychology, health care, marketing, education, etc., and can benefit both the individual and the society. The research here focuses on identifying the seven basic emotions (angry, disgust, happy, sad, fear, neutral, surprise) of the students in a B-School across four time slots during a day for 6 days (Monday–Saturday) using deep learning algorithm. To achieve this, performance of three deep learning models namely AlexNet, VGG16, and VGG19 are compared and based on the performance AlexNet is selected to detect the emotions of the students. Statistical techniques including 2-way ANOVA and Multiple Linear Regression are also utilized to validate the results. Results of analysis reveal that majority of the students go through two emotions: Happy and sad throughout the day with neutral following. Extreme emotions like anger, fear, and disgust are observed among very few students. Results also reveal that time slot and day are not significant in determining the emotions. The outcome of this research can be utilized by teachers and top management who are the two predominant stakeholders in the education system concerned with the well-being of the students.