Integrating Boolean Logic-Based Feature Representation and Artificial Neural Networks for Digital Divide Level Classification in Online Learning

Abstract

Digital divide remains a major challenge in online learning, affecting students’ ability to access and benefit from digital education. While previous studies have primarily focused on first-level digital divide factors or employed machine learning models using raw input variables, limited research has explored the integration of logical feature engineering and artificial intelligence to model the multidimensional nature of digital divide. This study proposes a hybrid classification framework that combines Boolean Logic-based feature engineering and Artificial Neural Networks (ANN) for digital divide classification in online learning. The proposed framework utilizes four variables, namely Primary Device, Internet Stability, Equity Score, and Accessibility Score. Boolean Logic operations (AND, OR, and XOR) were applied to generate additional logical representations of digital access conditions before the ANN classification process. Data were collected from 324 student respondents participating in online learning. The model was evaluated using accuracy, precision, recall, and F1-score metrics. Experimental results demonstrate that the optimized ANN model achieved 93% accuracy, 93% weighted precision, 93% weighted recall, and 93% weighted F1-score. Comparative analysis revealed that Boolean Logic features contributed additional logical representations of digital access patterns, although their impact on predictive performance was not consistently significant. The findings highlight that digital divide should be viewed as a multidimensional phenomenon and demonstrate the potential of ANN based predictive modeling to support data-driven digital inclusion strategies in education.