Real-Time IoT Framework for Data Continuity of Digital Shadows

Modern IoT systems rely on continuous data streams to support real-time monitoring, analytics, and decision-making processes. However, these systems are inherently subject to data disruptions caused by network instability, device failures, or resource constraints, leading to missing measurements and temporal gaps. This thesis presents a real-time IoT framework designed to ensure data continuity in the presence of such disruptions. The proposed approach integrates a scalable data ingestion pipeline with a predictive module capable of generating short-term forecasts to replace missing observations. The framework is evaluated through extensive experiments considering both system-level performance and forecasting accuracy. Results show that architectural optimizations improve scalability under high data rates, while forecasting models effectively preserve data stream continuity. A runtime-aware model selection strategy is introduced to balance prediction accuracy and computational latency, enabling the system to adapt to dynamic operational constraints. The proposed solution demonstrates how combining data engineering and predictive modeling techniques can enhance the robustness and reliability of real-time IoT systems.