TrafficFlowAI:Attention-LSTMforReal-TimeHighwayCongestion Prediction with an Interactive Route Planning Dashboard

This paper presents TrafficFlow AI, an end-to-end intelligent transportation system that predicts real-time highway congestion using a custom Attention-LSTM deep learning model trained on 48,204 hourly traffic records from the UCI Metro Interstate Traffic Volume dataset (I-94 highway, Minnesota, USA). The core model combines a Long Short-Term Memory (LSTM) network with a self-attention mechanism (71,362 trainable parameters) that dynamically weights a 24-hour input window to forecast traffic volume for the next hour, achieving a test MSE of 0.00378 — a 12% improvement over a standard LSTM baseline. The system is deployed as a full-stack web application: a FastAPI backend serves the trained model as REST endpoints, consumed by a React + Leaflet.js interactive dashboard that visualises live and predicted congestion across four simulated monitoring stations along the I-94 corridor with one-click route-level forecasting. This work contributes to UN Sustainable Development Goal 11 (Sustainable Cities) and SDG 13 (Climate Action) by enabling data-driven traffic management that reduces fuel waste and urban emissions. Keywords: LSTM, Attention Mechanism, Traffic Flow Prediction, Intelligent Transportation Systems, FastAPI, React.js, Leaflet.js, Deep Learning, SDG 11, SDG 13