Massive Learning Behaviours Influence Educational Sustainability: A Machine Learning Approach

Discovering learning behavior patterns while using Massive Open Online Courses (MOOCs) or e-learning platform education is being one of the substantial aspects in the plethora of past studies. In this study, we predict the taxonomy of learner behaviors by using the data of Junyi Academy, an e-learning platform similar to Khan Academy. We evaluate the dataset of first 2,000 unique learners and applying machine learning algorithms, i.e., Logistic Regression, and Neural Network (NNET) to predict the status of learners’ behavior. Our result shows that NNET autotuned is high accuracy champion predictive model as compared to Logistic Regression and NNET manually tuned hyperparameters models. The accuracy we measured based on Kolmogorov-Smirnov (KS) statistic for VALIDATION partition dataset. Furthermore, our model allows teachers to track or predict behavioral changes of the individual student and reveal insights into the content of the course material. It will also assist teachers in getting an early prediction of behaviors before finishing the course or skill, which is constructive and valuable for sustainable education or educational sustainability.