RNN-GRU Powered Multi-Level Traffic Classification in Smart Cities with IoV Sensor Data

Abstract

Recent studies show that urban traffic congestion costs the global economy nearly $300 billion annually, with delays in smart cities increasing by over 15% in the last five years despite Internet of Vehicles (IoV) adoption. This growing challenge underscores the need for intelligent, data-driven traffic classification systems to optimize flow and reduce congestion. However, existing IoVbased traffic prediction models often suffer from low adaptability to diverse city conditions and limited handling of complex, high-dimensional sensor data, leading to inaccurate classifications and suboptimal traffic control decisions. In this work, we present a novel IoV-driven Smart City Traffic Classification framework designed to accurately categorize traffic conditions into High, Low, Medium, and Very High levels using multi-source sensor data. The dataset incorporates diverse features, including City, Vehicle Type, Weather, Economic Condition, Day of Week, Hour of Day, Speed, Is Peak Hour, Random Event Occurred, Energy Consumption, Traffic Density, and Traffic Density Category. Our approach begins with robust data preprocessing, including missing value handling, normalization, and categorical encoding to ensure clean and standardized inputs. The processed data is then split into training and testing sets to enable unbiased performance evaluation. We benchmark traditional machine learning classifiers such as KNearest Neighbors (KNN) and Logistic Regression, before introducing our proposed Recurrent Neural Network with Gated Recurrent Unit (RNN-GRU) model. This architecture captures sequential dependencies in temporal and contextual traffic data, enabling more accurate classification under dynamic urban scenarios. Comparative analysis shows that the proposed RNN-GRU significantly outperforms baseline methods in precision, recall, and F1-score, offering a scalable and adaptive solution for real-time traffic management in smart cities.