2 of 7

CODE VERSE HACKATHON 2025

Problem Statement Title: Create a short-term prediction model for a next generation predictive traffic alert system.

Team Name: Spark AI

Team Members: Om Thorat, Ashwini Thakre, Pranav Thete, Nikita Tengle.

3 of 7

Proposed Solution: AI-Powered Predictive Traffic Alerts

We propose a short-term traffic prediction model using machine learning to forecast traffic congestion 5–30 minutes in advance. The model combines real-time and historical traffic data with algorithms like LSTM and Graph Neural Networks to generate accurate, localized predictions.

This solution helps users avoid delays by providing early alerts and alternative routes, reducing travel time and congestion. Its innovative use of predictive modeling and real-time data makes it more proactive and adaptable than existing reactive systems.

4 of 7

TECHNICAL APPROACH

Technologies to be Used:

Programming Languages: Python

Frameworks & Libraries: TensorFlow, PyTorch, Scikit-learn, Pandas, NumPy

Tools: Jupyter Notebook, Google Colab, Docker

Data Sources: Real-time traffic APIs (e.g., Google Maps, HERE), GPS/sensor data.

Methodology and Process for Implementation:

Data Collection: Gather real-time and historical traffic data.

Preprocessing: Clean, normalize, and structure the data.

Model Development: Train models (LSTM, GNN) for time and spatial prediction.

Evaluation: Use metrics like MAE, RMSE to assess performance.

Deployment: Deploy as an API or integrate into a mobile/web app for live alerts.

Testing & Feedback: Continuously monitor and update the model with new data.

5 of 7

FEASIBILITY AND VIABILITY

Feasibility Analysis, Challenges & Mitigation

Feasibility of the Idea: The idea is highly feasible due to the availability of real-time traffic data, advancements in machine learning (especially time-series and spatial modeling), and the growing need for intelligent traffic solutions in urban areas.

Potential Challenges and Risks:

Incomplete or noisy traffic data

Real-time system latency

Model accuracy in unpredictable situations (e.g., accidents, weather)

Data privacy concerns

Strategies to Overcome Challenges:

Use robust data-cleaning techniques and multiple data sources to improve reliability.

Optimize model performance with lightweight architectures for real-time use.

Include anomaly detection modules for unexpected events.

Ensure compliance with data privacy regulations through anonymization.

6 of 7

IMPACT AND BENEFITS

Potential Impact: The solution can significantly improve urban mobility by predicting and reducing traffic congestion. It supports smarter city planning, lowers environmental impact, and enhances the commuter experience through real-time, predictive alerts.

Benefits of the Solution:

Reduced travel time and fuel consumption

Lower carbon emissions due to optimized routes

Improved emergency response and logistics planning

Enhanced user satisfaction through proactive traffic alerts

Scalable solution adaptable to different cities and road networks

7 of 7

THANK YOU

Questions & Discussion

Spark AI