NYC Transit Pulse

• 

  Home Page

Inspiration

We were inspired both by our interest in public transportation. We built NYC Transit Pulse to transform raw MTA data into actionable predictions, empowering riders to make smarter travel decisions.

What it does

NYC Transit Pulse predicts bus delays in real-time using machine learning by analyzing historical patterns, using time of day, what bus stop the user is at, and the day of the week.

How we built it

1. Data Pipeline:
   • Processed 18M+ MTA records using Python/Pandas
   • Engineered temporal features (rush hour, weekend flags)
2. Machine Learning:
   • Trained a Huber Regressor model (scikit-learn)
3. Deployment:
   • Flask backend with REST API
   • React.js frontend for real-time updates ## Challenges we ran into
4. Data Quality: 23% of MTA records had missing/inconsistent stop IDs
5. Cold Starts: Predicting delays for new routes with no history ## Accomplishments that we're proud of
6. Our model's accuracy has improved drastically after using different features ## What we learned
7. Scikit-learn pipelines enable rapid experimentation ## What's next for NYC Transit Pulse
8. Adding subway/train delay predictions

Built With

• flask
• mta-bus-time
• python
• scikit-learn