Process of building a dataset and classification of vark learning styles with machine learning and predictive analytics models

Abstract

As there is a rise in the online and customized learning platforms, learning style preferences give us insight into better utilization of educational resources available. VARK learning styles are developed by Fleming and Bonwell on the premises of preferred intake of information by the students. VARK model describes four sensory modalities Visual, Kinesthetic, Read/Write and Auditory respectively for the input. Initially learning styles are calculated using the VARK questionnaire as an instrument. In this case machine learning and predictive analytics can help classify learning style by including all the descriptors including demographic and behavioural. This study will explore the data mining process from the raw data collected from the college students through a questionnaire which is an integral part before the analysis to determine possibility to use different descriptors. This study primarily aims to classify the students based on VARK learning styles based on their sensory modalities and build a predictive model using the variables that can influence the learning styles. This study explored the relationship between demographic factors like school and place people grew up. Results proved to contradict those factors. We concluded that with the growth of big data learning style classification, a blend of model algorithms or stacked algorithms like voting classifier can be used to adapt to a user application.