Predicting Student Outcomes through Cognitive Behaviour Analysis in Online Collaborative Learning Using Big Data and Machine Learning

Background: By bringing together geographically separated students, collaborative learning (CL) technologies such as Google Classroom and Skype for Education have revolutionized digital education. Using Data Analytics (DA) and Machine Learning (ML) models, this study assesses how well CL techniques improve cognitive engagement and forecast academic outcomes. Methods: Analysis was done on a dataset of 5,000 student records from an open-access CL platform. Peer feedback scores, quiz results, assignment submission rates, and interaction frequency were among the parameters. ML models were trained, including Factorization Machine (FM) classifiers, Random Forest, Decision Trees, and Support Vector Machines. Accuracy, precision, recall, and F1-score were used to evaluate performance, and FM was contrasted with Gold Standard classifiers such as logistic regression. Findings: The FM classifier outperformed Random Forest (86.4%) and Logistic Regression (82.5%) with the greatest prediction accuracy of 91.8%. The two most important factors, peer feedback and interaction frequency, helped to enhance the models by 22% over the baseline. Novelty and applications: This is the first extensive study that integrates multifactorial engagement characteristics for academic prediction and applies FM classifiers to CL data. It presents new avenues for the study of personalized learning analytics and validates FM-based DA models as useful instruments for adaptive, real-time educational support. Keywords: Collaborative Learning, Big data, Machine Learning, Student academic performance, Cognitive behaviour