The Impact of Behavioural Features on Predicting Academic Success: A Machine Learning Approach

Abstract-To discover hidden patterns from educational data, researchers are developing methods by using educational data mining. Dataset and its features/attributes determine the eminence of data mining techniques. Student’s academic performance model by using a new class of features i.e., behavioural features was built in this research paper. These are significant features as they are associated with the learner interactivity in e-learning system. Data was collected from an e-Learning system called Kalboard 360 using Experience API web service called (xAPI). After data preprocessing and feature selection, machine learning algorithms such as Decision Tree, Support Vector Machine and Artificial Neural Network were used to build the model. It is clearly visible from the results that there is a sturdy association between learner behaviours and its academic achievement. Results with above-mentioned classification methods using behavioural features attained up to 25% enhancement in the accuracy as compared to the results when same classification methods were applied on the data set without behavioural features.