<p>Real-time applications are rapidly evolving with the Internet of Things (IoT), as smart environments and sensor networks continuously generate massive amounts of data. Effective data management, covering data generation, storage, processing and energy use, is vital for IoT systems operating under resource constraints. This paper proposes a lightweight data management framework that consists of three IoT data models viz. a data generation model, an accuracy-oriented data model, and a performance-oriented data model. By using dynamic polynomial models for data segments, the framework achieves efficient storage and real-time processing while minimizing redundancy. The proposed framework is evaluated with real-world environmental data, including temperature, humidity, and pressure readings collected from IoT sensors. Results show that the accuracy-oriented data model achieves up to 96% data reduction with mean errors of 0.0372, while the performance-oriented model attains data reductions of up to 98% with a mean error of 0.0761. The proposed data models are compared with a typical IoT scenario in terms of energy consumption, processing time, storage space, and network lifetime. Findings confirm that the proposed data models optimize resource usage, making them suitable for real-time, resource limited environments. To further validate the proposed data management framework, a comparative evaluation with existing works is conducted in terms of accuracy, precision, recall, storage and energy efficiency, and processing time. Results indicate that the accuracy-oriented data model achieves 96.5% accuracy, 95.8% precision, and 96.1% recall with high energy savings, while the performance-oriented data model provides 98% storage efficiency, 0.72s processing time, and 64.1 mWh energy use. Overall, the framework offers a promising solution for real-time IoT deployments with limited resources.</p>