Machine Learning-Based Early Detection of Learning Disabilities in School-Aged Children

Abstract

Early identification of learning disability in school age child has been a major challenge because of the delayed diagnosis, inadequate screening instruments and inconsistency in the cognitive and behavioural patterns. This research solves the issue by suggesting a machine learning-based model of timely and accurate identification of learning disabilities using academic, behavioral and cognitive assessment information. The main goal is to come up with a predictive model that aids better strategies of early intervention and better educational achievement. It uses three machine learning techniques, namely, Support Vector Machine, Random Forest, and Artificial Neural Network. The algorithms get trained using structured data sets of student performance measures, attention measures, and psychometric measures. The parameters used compared to assess the accuracy, precision, recall, F1-score, and area under the curve (AUC) are used to compare them. Findings show that the Artificial Neural Network performances better than the other models with a accuracy of 94.2, then Random Forest with a 91.6 and lastly Support Vector Machine with 88.9. The ANN is also better in recall, which means that it recognizes at-risk students better. Random Forest will be more reliable in terms of its performance with less variance whereas SVM will be effective in its baseline classification. The research area is applicable to school-based screening, real-time educational monitoring systems. The suggested strategy helps teachers and practitioners to make evidence-based decisions to intervene early. Finally, machine learning methods are important to enhance the accuracy of early detection and proactively support children with learning disabilities, which will help to build an inclusive education system.