TRAFFIC PREDICTION USING MACHINE LEARNING

In modern urban environments, efficient traffic management is a key element of smart city development. Withgrowing vehicle populations and increasingly complex traffic flow patterns, the demand for accurate and realtime traffic prediction systems is more crucial than ever. This project presents a robust traffic prediction frameworkthat utilizes the YOLOv8 (You Only Look Once version 8) deep learning model in conjunction with OpenCV toanalyse real-time traffic conditions and forecast traffic density effectively. The system accepts input from varioussources including static images, video recordings, and live webcam feeds. It performs object detection usingYOLOv8 to count vehicles and evaluate traffic density across time intervals. These traffic metrics are then usedto predict short-term traffic flow trends, offering insights that can aid in congestion control and dynamic trafficsignal management. YOLOv8's efficiency and high accuracy make it ideal for processing visual data in real-timescenarios. Key features of the system include confidence-based detection filtering, real-time frame annotation,Fullscreen visual output, and persistent storage of analysed data. The collected traffic data serves as the basis forpredictive modelling, which can be extended using time- series forecasting algorithms such as ARIMA or LSTMfor more advanced traffic forecasting. This project showcases how integrating deep learning with intelligent videoanalysis can provide scalable and real-time traffic prediction, ultimately supporting smarter and safer urbanmobility solutions.