A Hybrid CNN-LSTM Model for Predicting Learning Outcomes in Students with Autism Spectrum Disorder

Abstract

Autism Spectrum Disorder (ASD) students often struggle with adaptive learning, behavioral engagement and academic performance prediction because of heterogeneous cognitive pattern and limited educational analytics to provide personalized analytics.Students with Autism Spectrum Disorder (ASD) have a tendency to have heterogeneous cognitive pattern and limited educational analytics to provide personalized education analytics and therefore struggles with adaptive learning, behavioural engagement, and academic performance prediction. This research introduces a novel multimodal CNN–LSTM model to predict learning outcomes for students with ASD, by leveraging their educational and behavioral data. The method combines Convolutional Neural Networks (CNN) to learn spatial and behavioural features, and Long Short-Term Memory (LSTM) networks to learn the temporal features of student interaction sequences and performance progression. The model proposed (Hybrid NeuroLearn-ASD) was evaluated based on the publicly available Autism Screening Adult Dataset, in conjunction with learning activity data and cognitive assessment data indicators. Comparative analysis was done with Support Vector Machine (SVM), Random Forest (RF), standalone CNN and standalone LSTM models of the traditional Machine Learning approach. Experimental results showed the proposed hybrid CNN-LSTM model outperformed the existing models by 5–10% in terms of prediction accuracy, precision, recall, F1-score, and ROC-AUC scores, which were 96.4%, 95.9%, 95.2%, 95.5% and 97.1%, respectively. The framework greatly enhanced the adaptive educational prediction, early intervention capability and personalized learning assessment for ASD students. The key novelty of this research is to combine the extraction of spatial-behavioral features and the sequential cognitive learning analysis to achieve intelligent prediction of educational outcomes. Proposed model is an efficient AI-based solution for inclusive and personalized special education system.