Streamlined Traffic Prognosis using Flexible Reservoir Sampling and Regression Methods

The increased use of Internet of Things (IoT) technologies has resulted in an exponential increase in real-time data streams, particularly in smart city applications, especially for traffic management. Accurate prediction of traffic parameters in such environments is critical for optimizing traffic flow, reducing congestion, and enabling efficient resource management. This study presents an approach to the prediction of traffic intensity and occupancy using IoT streaming data, time series analysis, and machine learning algorithms. The proposed method includes preprocessing steps such as data interpolation to handle missing values and temporal alignment, followed by feature extraction and model training using a combination of regression and sampling techniques. Experiments were carried out on a real-world IoT traffic dataset, and the results show significant improvements in the prediction accuracy in terms of MAPE values. It also predicts the complex event of congestion, using a rule-based algorithm. The proposed method can pave the way for smarter and more efficient urban infrastructure.