Exploring the VAK model to predict student learning styles based on learning activity

• Research goal. • Development of a combined system that suggests students’ appropriate evaluation methods and learning preferences based on all their academic data. • To achieve this purpose, two phases are used. In the first phase, there are two key steps that each employ a different methodology. To use the VAK learning style Model, the learner input must first be mapped using semantic mapping techniques (VAK stands for (V) visual, (A) auditory, and (K) kinesthetic). The second phase involves assigning students to select learning styles from the three VAK learning styles based on the data from their learning activities. We predict a suitable evaluation technique according to the preferred learning style for students. • We use a dataset (open university learning analytics dataset) as a learning benchmark dataset to evaluate the efficiency of our new model. • Our proposed model can categorize students with an average accuracy of 98 % based on their various learning activities and forecast the appropriate evaluation methods for each student. Adaptive learning systems focus on improving the performance of educational processes by adapting them to different students. One of the factors which require this adaptation is the preferred way of students to learn, which is at times considered as a blend of visual, auditory, kinesthetic, (VAK) etc. Knowing such things, not only helps the teacher to improve the delivery of the content, but also assists in improving assessment methods to suit each student. The primary motivation of this research is to analyze students’ engagement characteristics in Virtual Learning Environments (VLE) and determine their prevalent instructional preference and learning style and recommend the best learning assessment tools. To accomplish this goal, we have proposed an integrated system which encompasses the use of machine learning (ML) algorithms. This hybrid model is aimed at linking various activities to VAK model of learning and hence place students in their various class learning preferences derived from their activities and the patterns created during the learning processes. We used the Open University Learning Analytics Dataset (OULAD)to assess the efficiency of the proposed system. Multiple tests were performed by different machine learning classifiers, mainly in predicting learning style and recommending an assessment methodology. Our results show that the Random Forest algorithm achieved the highest accuracy with 98 %.This research shows how machine learning techniques embedded in learning analytics could expand the functionalities of VLEs toward greater personalization and effectiveness, with every student receiving the best educational experience that suits their learning styles.