Modeling and Predicting Online Learning Activities of Students: An HMM-LSTM based Hybrid Solution
Alexis Amezaga Hechavarria, M. Omair Shafiq · 2021 20th IEEE International Conference on Machine Learning and Applications (ICMLA) · 2021
With significant increase in popularity of online education, need for educators to get to know about learning experiences of students and to provide feedback to students has also been increasing. In this paper, we propose an HMM-LSTM based hybrid solution to model and predict online learning activities of students using online learning management systems (LMS) or platforms and provide real-time feedback to students. Our solution is a smart classifier empowered by and based on the Markov-Chain (MC) approach, specifically a Hidden Markov Model (HMM), with the Long Short-Term Memory (LSTM) neural network. The novelty is in the use and treatment of hidden data and metrics to raise flags that indicate outlier online student behavior based on historical data from the same online session and other sessions in the past. We propose a design of a system in which we utilize HMM and LSTM, and the LSTM component of the model is ‘advised’ by the HMM probability used in the metrics that drive the outlier detection process. The system relies on the LSTM prediction to perform early detection of patterns.