A digital literacy predictive model in the context of distance education

Abstract

Digital technologies are essential in the distance education environment. Several learners who enrol in distance education programs have difficulty using the technologies during the learning sessions. This study focuses on factors influencing digital literacy in distance education programs. A predictive model was developed based on four phases of the study; preliminary study, dataset preparation, experimental design, and analysis. For this study, 232 survey data were successfully collected. The experimental design has two phases, including data pre-processing and feature selection. - The target class is defined based on the device used and the generation. Five machine learning algorithms have been selected for the classifier model in the data analysis phase. They are logistic regression as a baseline model, k‐nearest neighbour, random forest, support vector machine (SVM), and multilayer perceptron. In addition, five cross-validation settings were chosen, which are 2, 3, 5, 10, and 20‐fold cross‐validation. The comparison result for two‐target data will be based on accuracy, precision, recall, and F-measure. Using the SVM algorithm, the best model shows the maximum accuracy with 82.9% for model learning based on generation and 41.4% for model learning based on the device. This study identifies interesting patterns based on learners’ generation. It shows that different generations have various ways of using digital literacy. Thus, the proposed predictive model proves that generation is a useful indicator to recognize the level of digital literacy among higher education learners, especially those for distance learning.