Chronos-Traffic: An AI-Driven Framework for Spatio-Temporal Traffic Forecasting and Anomaly Detection
Year: 2026
Type: preprint
Abstract: Chronos-Traffic presents an AI-driven framework for spatio-temporal traffic analysis, forecasting, and road anomaly detection. The system combines historical traffic data analysis using regression-based forecasting models with a MobileNetV2-based convolutional neural network (CNN) for classifying road conditions into normal traffic, construction and accident categories. The framework integrates a FastAPI backend with a Streamlit dashboard to provide traffic visualization, forecasting results, model performance evaluation and image-based anomaly detection through a unified interface. Experimental evaluation demonstrates the system's classification performance on a held-out dataset of 819 images. The study also evaluates Ridge Regression against Random Forest and Persistence baselines for traffic forecasting, discussing the results and current limitations. This work presents an integrated prototype for intelligent traffic monitoring and highlights opportunities for future improvements through larger datasets, independent validation, and more advanced spatio-temporal forecasting techniques.
Author: Babar, Shaheer
Institution: Bahria University
Language: English
Subfield: Transportation
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7225626277
Cites:
Cited by:
Related to: