UrbanFlow360: A Cloud-Native Architecture for Real-Time Urban Traffic Congestion Prediction

Abstract

UrbanFlow360 is a cloud-native Software-as-a- Service (SaaS) platform for real-time traffic congestion prediction and urban traffic analytics. Developed using a modular mi- croservice architecture and containerized with Docker, the system integrates predictive machine learning pipelines and offers live analytics through an interactive Streamlit dashboard. This paper details the end-to-end development lifecycle of UrbanFlow360, including dataset preprocessing, model training, application de- velopment, containerization, and deployment on AWS Elastic Container Registry (ECR) and Amazon EC2. The architecture demonstrates best practices in reproducibility, portability, and scalability of intelligent transportation systems.

Introduction

UrbanFlow360 tackles urban mobility challenges caused by increasing vehicle numbers, infrastructure limits, and unpredictable congestion. Traditional traffic systems lack predictive capabilities, which UrbanFlow360 addresses by leveraging cloud-native technologies for real-time congestion prediction via a browser-based dashboard built with Streamlit and deployed on AWS.

Team Roles

• Akash Tiwari: Cloud and DevOps (EC2, Docker, ECR, IAM)

• Aashish Dewangan: Data preprocessing and ingestion (CSV, JSON, schema design)

• Kishan Kumar Bouri: Machine Learning pipeline and training

• Yash Mathur: Frontend integration and UI/UX improvements

Data & Preprocessing

• Data sourced from traffic logs in Delhi and Bangalore.

• Preprocessing includes:

  • Missing value imputation

  • Time and weekday feature extraction

  • Encoding categorical variables and scaling

System Architecture

• Modular design with components for data preprocessing, model inference, UI, and cloud deployment.

• Model uses a Random Forest Regressor for robust, interpretable predictions.

Model Performance

• Evaluated with:

  • Mean Absolute Error (MAE)

  • Root Mean Square Error (RMSE)

  • R² score

• Achieved effective accuracy on city-level congestion forecasting.

Frontend & User Interaction

• Users upload traffic data and select cities via a Streamlit dashboard.

• Real-time congestion predictions are displayed with interactive charts.

Deployment

• Packaged as a Docker container for consistency.

• Deployed on AWS EC2 instances via ECR with proper IAM roles.

• Accessible through EC2’s public IP address.

Results Sample (Bangalore)

UrbanFlow360 offers a scalable, cloud-powered solution for predictive traffic management, improving urban mobility through timely insights and user-friendly visualization.

Conclusion

UrbanFlow360 bridges cloud computing and urban mobility analytics. It is portable, scalable, and accessible to both cities and researchers. Future work includes multi-city support, live APIs, and SUMO integration for simulation.

References

[1] DLR SUMO - https://sumo.dlr.de/docs/ [2] AWS Docs - https://docs.aws.amazon.com/ [3] Streamlit - https://docs.streamlit.io/ [4] Scikit-learn - https://scikit-learn.org [5] Docker Docs - https://docs.docker.com/ [6] GitHub Repo - https://github.com/KishanBouri/urbanflow360

Copyright

Copyright © 2025 Akash Tiwari, Aashish Dewangan, Kishan Kumar Bouri, Yash Mathur. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.