MIRERC086/2025: AI-DRIVEN CIRCULAR ECONOMY MODEL FOR E-WASTE SORTING BASED ON ADAPTIVE LEARNING

Abstract

The growing volume of e-waste presents significant environmental and health challenges, particularly in developing regions lacking sustainable management systems. This study develops an AI-driven circular economy model for e-waste sorting, integrating Explainable Artificial Intelligence (XAI)—which provides transparency into model decisions—and adaptive learning techniques to enable continuous improvement. Using a mixed-methods approach, the research combines qualitative insights from stakeholder interviews with quantitative analysis of annotated e-waste image datasets. The technical implementation features a fine-tuned EfficientNet-B4 model enhanced with Grad-CAM++ for visual explanations and Elastic Weight Consolidation to support adaptive learning. An accompanying mobile application supports geolocation-based collection points and real-time categorization of e-waste. The proposed model undergoes rigorous testing to ensure accuracy and interpretability, while user trust metrics are assessed through descriptive analysis, and stakeholder feedback is thematically analyzed. The results aim to advance Kenya’s e-waste policy framework and provide a scalable prototype adaptable to both public and private sector contexts, demonstrating how AI can enhance efficiency, transparency, and sustainability in circular economy approaches to e-waste management