A Hybrid Deep Learning-Enabled Meta-Learning Framework for High-Accuracy Multi-Domain Prediction Systems

Abstract Deep Learning (DL) and Machine Learning (ML) techniques have achieved significant progress in domains such as healthcare diagnostics, financial forecasting, and intelligent transport systems. However, traditional DL models struggle to generalize across diverse environments, requiring large labeled datasets and frequent retraining. Meta-learning offers a solution by enabling models to rapidly adapt to new tasks with minimal data. This paper proposes a Hybrid Deep Learning-Enabled Meta-Learning Framework (HDLM-LF) designed to enhance multi-domain prediction accuracy through a combination of Convolutional Neural Networks (CNNs), Transformers, and Meta-Agnostic Meta-Learning (MAML). The framework is evaluated across healthcare, financial, and IoT datasets, demonstrating improved prediction accuracy, responsiveness to real-time adaptation, and computational speed. The performance metrics are summarized in Tables 2–4. The study shows that HDLM-LF outperforms existing meta-learning and deep learning techniques in accuracy while reducing model retraining and latency.