AI-Based Multi-Domain Early Risk Prediction System for Learning Difficulties in Students

Abstract

The timely detection of learning difficulties in students is a fundamental necessity for effective intervention. In this paper, we propose a novel AI-driven multi-domain risk prediction system called EarlyPredict. The system is developed to identify students at risk of dyslexia, attention deficit, cognitive processing disorder, and poor academic achievement. The system is developed by training four individual Random Forest classifiers using two publicly available datasets: the Open University Learning Analytics Dataset (OULAD) and a dataset related to behavioral dyslexia interaction. The decision fusion engine is developed by implementing a rule-based system that aggregates individual predictions from each academic, dyslexia, attention, and cognitive domain to produce a single risk level classification: Low, Moderate, or High. The accuracy of individual models is found to be up to 99%. The proposed system is a significant advancement in the detection of learning difficulties in students.