NYC Ride Demand Forecasting (2015 - 2024)

A full end-to-end project forecasting hourly NYC yellow taxi demand using 10 years of trip data, weather conditions, NYC events, and rich time-based features. Built to demonstrate real-world demand forecasting, time-series modeling, and marketplace analytics skills.

---

Live Dashboard

Explore predictions interactively: Live Dashboard

---

Project Summary

This project analyzes and forecasts how NYC ride demand evolves with:

• Daily + weekly seasonality (commute cycles, weekend behavior)
• NYC permitted events (concerts, sports, parades)
• Weather factors (temperature, precipitation, wind)
• Calendar dynamics (holidays, weekdays, month patterns)
• Long-term structural changes, including the COVID demand collapse

The final model uses XGBoost, producing accurate and stable short-horizon forecasts.

---

Key Insights

Demand Behavior

• Clear two-peak weekday pattern (AM/PM commute).
• Fridays have the highest demand; Mondays the lowest.
• Events noticeably elevate ride volume.
• Weather impacts demand moderately (rain increases rides slightly).
• Summer months show lower demand than winter.
• A major structural break occurs in 2020 due to COVID.

Model Performance

Models evaluated:

• Naive & seasonal naive
• Linear regression
• Random Forest
• XGBoost (best)

XGBoost captured peak timing and magnitude well, achieving low MAE/RMSE and strong generalization to unseen weeks.

More detailed analysis and visualizations are available in the modeling notebook.

---

Notebooks

Each notebook contains visualizations, insights, and technical explanation:

• 1_eda.ipynb - Trend, seasonality, weather, events, structural break
• 2_feature_engineering.ipynb - Lags, rolling features, calendar encoding
• 3_modeling.ipynb - Baselines, models, evaluation, final metrics

Each notebook provides deeper insight and commentary.

---

Running the Dashboard Locally

pip install -r requirements.txt
streamlit run dashboard/app.py

Everything works immediately - processed data and the trained model are included.

---

(Optional) Rebuild the Dataset From Scratch

If you want to reproduce the full data pipeline using raw TLC data + APIs:

python -m src.taxi.ingest
python -m src.weather.fetch_weather
python -m src.events.fetch_events
python -m src.events.expand_events
python -m src.features.build_base_table
python -m src.prepare_model_data

This regenerates the same processed files included in the repository.