Context-Aware Prediction Models for Dynamic Data Streams

Abstract

Context-aware intelligence enhances predictive systems by integrating situational dependencies and temporal variability in complex environments. However, many methods inadequately address continuous streaming traffic dynamics and evolving contextual interactions in multiple scenarios. This research aims to develop a dynamic context-aware prediction model for streaming traffic data using a deep neural network to improve adaptive trajectory and flow prediction. Traffic data is aggregated from instant sensors, vehicular networks, and open traffic repositories, ensuring diversity in spatial–temporal patterns. Pre-processing employs Z-Score normalization for cleaning and standardizing the data. Feature extraction utilizes Linear Discrimination Analysis (LDA) to capture dominant mobility patterns. The model systematically acquires streaming inputs, performs sequential cleaning, extracts compact representations, and feeds them into a context-aware predictive engine. The proposed Vortex Algorithm-driven Dynamic Deep Neural Network (VA-DDNN) integrates Vortex Optimization for adaptive weight tuning and dynamic hyperparameter adjustment, enabling efficient convergence under fluctuating traffic conditions. The DDNN component models non-linear temporal dependencies, while the Vortex Algorithm enhances learning stability and avoids local minima during continuous updates. The model incorporates dynamic context modeling by capturing vehicle interactions, traffic density transitions, and environmental variations, enabling responsive prediction under streaming conditions. The Performance demonstrates improved prediction consistency, reduced latency, enhanced adaptability, and robustness across varying traffic scenarios with a precision of 97.4%, a recall of 92.5%, an F1-score of 94.8%, and a mAP of 92.10%. The method establishes an efficient model for the context-aware traffic prediction, supporting intelligent transportation systems with scalable and adaptive decision-making capabilities.