Development of analytics tools for e-Learning

Abstract

Students are constantly classified based on factors such as their personal backgrounds, grades, and their behaviour in school. With the rise of data analytics and machine learning, such classifications are now extended into the form of prediction models. Machine learning applications have become increasingly common in education today due to the sheer amount of information available through online systems. Currently, we want to find out whether there are any links between student background, e-Learning usage, and student performance. In this project, we are aiming to build a prediction model based on based student background data, weekly e-Learning, and student class performance. The main objective is find out how well final grades can be determined before the final examination. Furthermore, we also want to understand the impact of different background factors and e-Learning on the students’ performance so that useful and actionable insights can be obtained. Classification algorithms such as K-Nearest Neighbours, Support Vector Machines, Decision Trees, and Random Forest, covering linear, non-linear, and ensemble methods were used in this project. Various prediction models were developed and tested, and we were able to predict students’ grades with a maximum mean accuracy of 56% near the end of the course. Through this project, we were also able to identify trends based on background factors on student performance to complement the insights with the prediction model so that the results and findings of this project can be implemented in future classes.