System Design for Accurate Urban Traffic Classification with Machine Learning: Findings from KNN, SVM, and Decision Tree

Abstract

Managing traffic in cities is essential to public safety, reducing congestion, and improving the flow of vehicles on busy streets. With the increasing size and complexity of urban areas, effective traffic management systems require dynamic, adaptable solutions. This study combines machine learning models and Raspberry Pi systems for real-time traffic classification and monitoring. Three machine learning models K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Decision Tree Classifier were evaluated for classifying traffic levels as “Low” or “High” using the City Plus dataset. The Raspberry Pi system was employed as a lightweight edge computing device for data acquisition and preprocessing. Among the models, the Decision Tree Classifier achieved the highest accuracy at 100%, followed by the SVM model at 98% and the KNN model at 96%. The integration of a Raspberry Pi system demonstrates the feasibility of a cost-effective, scalable approach to intelligent traffic management, enabling real-time insights and data-driven decision-making for smart cities.