Capstone Final Year Project — AI-Driven Transit Management

SmartDTCAI-Powered Transit - For Delhi People

Transforming Delhi Transport Corporation with LSTM demand forecasting, XGBoost delay prediction, and Genetic Algorithm scheduling — delivering smarter, faster, more reliable bus services for 3.5 million daily passengers.

Active Routes

Live Buses

Drivers On-Duty

Daily Passengers

Proven System Outcomes

Compared against traditional static scheduling baseline

78%

OTP Improvement

vs 62% baseline

47%

Wait Reduction

average savings

+18%

Fleet Utilisation

vs static plan

8.3%

Demand MAPE

LSTM accuracy

Core AI Components

Four machine learning models working together in a unified microservice architecture

LSTM Demand Forecasting

Deep learning predicts passenger demand 24 hours ahead with 91.7% accuracy across all 569 DTC routes.

XGBoost Delay Prediction

Machine learning forecasts bus delays with 4.2 min RMSE using weather, traffic and historical patterns.

Genetic Algorithm Scheduling

Evolutionary algorithm optimises headways and fleet allocation, reducing passenger wait times by 47%.

Real-Time GPS Tracking

Live bus positions streamed via Socket.io every 10 seconds, with automatic alert generation for delays.

Anomaly Detection

Isolation Forest model detects route anomalies and overcrowding, triggering instant fleet adjustments.

Arrival ETA Predictions

Gradient Boosting model predicts stop-level ETAs with 2.8 min MAE for the passenger mobile app.

Explore SmartDTC

Track buses live, plan your journey, or dive into the admin analytics dashboard.