Statistical Investigation of Student Behaviour Analysis Models From An Empirical Perspective

Student behaviour analysis is a multidisciplinary field which requires exploration of a wide variety of data, including, student’s geographical profile, area of behavioural study, temporal responses, situational responses, analytical reasoning, attention profile, etc. Combination of these factors requires design of intelligent machine learning approaches, which work on temporal behavioural responses. For instance, to predict student’s inclination towards technical education, models utilize analytical questionnaire, and social media tools to capture student’s behaviour. This data is processed using various deep learning architectures to estimate student’s inclination probability towards technical education. A wide variety of architectures are proposed for this task, and these architectures vary in terms of performance metrics, area of application, geography of student, etc. This makes it uncertain for researchers to test, validate &select most optimum models for their application, which increases cost & time needed for deployment. In order to reduce the uncertainty of model selection, this paper reviews some of the recently proposed methods for student behaviour analysis, and compares them in terms of performance metrics, area of application, and geographical parameters. The performance metrics include accuracy of analysis, computational complexity, mean squared error (MSE), and speed of analysis. This review will be helpful for researchers & behavioural analysis system designers to select the most optimum models for newer deployments, and will assist in performance upgradation of existing systems. Moreover, this text also recommends various improvements & enhancements in the reviewed models, which assists in upgrading their internal capabilities including scalability, flexibility, and performance analysis