Abstract:

Learning disabilities pose significant challenges to educational outcomes and personal development, impacting individuals’ academic trajectories and broader life opportunities. While early detection of such challenges is critical, existing screening processes often rely on manual methods, which can be time-consuming and prone to subjective interpretation. Recent advances in machine learning offer promising avenues for automated detection, potentially identifying at-risk individuals with higher precision and reliability. In this paper, we explore a dataset of 6,607 student records containing both cognitive and contextual information—such as hours studied, attendance, previous scores, and demographic factors—to build a classification model that predicts the presence of learning disabilities. We employ a variety of machine learning approaches, including Logistic Regression, Random Forest, XGBoost, and CatBoost. Hyperparameter tuning further refines the models, yielding an overall accuracy above 91% and demonstrating encouraging recall, precision, and F1-scores in different scenarios. Our findings suggest that integration of machine learning classifiers into educational settings may accelerate and improve the screening process for learning difficulties, leading to timely intervention. We also provide an in-depth discussion of feature importance, model comparisons, and recommendations for real-world implementation.

Keywords:

Learning Disabilities, Machine Learning, Classification, Random Forest, XGBoost, CatBoost, Educational Data Mining, Early Detection, Hyperparameter Tuning, Predictive Analytics.