Millennial Filipino Student Engagement Analyzer Using Facial Feature Classification

International Conference on Information Technology and Digital Applications 2017

计算机科学

Manseras, R.^1 ; Eugenio, F.^2 ; Palaoag, T.^3

University of the Immaculate Conception, Davao City, Philippines^1

Isabela Colleges Incorporated, Cauayan City, Isabela, Puerto Rico^2

University of the Cordilleras, Baguio City, Philippines^3

关键词: Affective Computing;    Classroom environment;    Conceptual frameworks;    Detection framework;    Emerging technologies;    Instruction delivery;    Naive Bayesian algorithms;    Student engagement;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/325/1/012006/pdf
DOI  :  10.1088/1757-899X/325/1/012006

学科分类：计算机科学（综合）

来源: IOP

【 摘 要 】

Millennials has been a word of mouth of everybody and a target market of various companies nowadays. In the Philippines, they comprise one third of the total population and most of them are still in school. Having a good education system is important for this generation to prepare them for better careers. And a good education system means having quality instruction as one of the input component indicators. In a classroom environment, teachers use facial features to measure the affect state of the class. Emerging technologies like Affective Computing is one of today's trends to improve quality instruction delivery. This, together with computer vision, can be used in analyzing affect states of the students and improve quality instruction delivery. This paper proposed a system of classifying student engagement using facial features. Identifying affect state, specifically Millennial Filipino student engagement, is one of the main priorities of every educator and this directed the authors to develop a tool to assess engagement percentage. Multiple face detection framework using Face API was employed to detect as many student faces as possible to gauge current engagement percentage of the whole class. The binary classifier model using Support Vector Machine (SVM) was primarily set in the conceptual framework of this study. To achieve the most accuracy performance of this model, a comparison of SVM to two of the most widely used binary classifiers were tested. Results show that SVM bested RandomForest and Naive Bayesian algorithms in most of the experiments from the different test datasets.