Visualizing the T: MBTA Rider Satisfaction Score

About

An interactive analytics platform that computes a comprehensive Rider Satisfaction Score (RSS) for the Massachusetts Bay Transportation Authority (MBTA) by combining operational performance metrics with passenger survey data.

Team Members: Dhanrithii Deepa, Eric Fu, Yuchen Cai
Live Demo: MBTA RSS Streamlit App

---

Project Goal

Develop a data-driven Rider Satisfaction Score that:

• Reflects both perceived experience (survey data) and operational performance (travel times, restrictions, crowding)
• Enables equity analysis across demographic groups and service areas
• Provides actionable insights at station/line/time-period granularity for MBTA service planning

---

How to Run

Prerequisites

• Python 3.9+
• pip or conda

Installation

# Clone repository
git clone https://github.com/urverse/Visualizing-the-T.git
cd Visualizing-the-T

# Install dependencies
pip install -r requirements.txt

Run Analysis Notebooks

# Start Jupyter
jupyter notebook

# Run notebooks in order:
# 1. notebooks/01_data_exploration.ipynb
# 2. notebooks/02_indicator_calculation.ipynb
# 3. notebooks/03_rss_computation.ipynb
# 4. notebooks/04_advanced_analytics.ipynb
# 5. notebooks/05_statistical_validation.ipynb

Run Streamlit Dashboard

streamlit run app/streamlit_app.py

---

Methodology

Data Sources

1. MBTA 2024 Passenger Survey - Rider satisfaction ratings and demographics
2. Rapid Transit Travel Times - Operational performance metrics
3. Fall 2024 SDP Ridership - Exposure weights by station/line/time
4. Speed Restrictions by Day - Service disruptions (slow zones)

Approach

1. Indicator Calculation

• Travel time reliability (P90/median ratio)
• Travel time volatility (coefficient of variation)
• Speed restriction coverage (% of line restricted)
• Crowding levels (ridership/capacity ratio)
• Actual survey satisfaction scores

2. RSS Computation

• Literature-based weights applied to normalized indicators
• Ridge regression validates weights against survey satisfaction
• Exposure-weighted aggregation using ridership data

3. Statistical Validation

• ANOVA hypothesis testing for line comparisons
• Bootstrap confidence intervals (95% CI)
• Equity analysis across income/race demographics

Key Findings

• Orange Line achieved highest RSS (73.2) despite serving predominantly low-income riders
• Methodology is equity-neutral - no systematic bias against underserved communities
• Reliability is strongest satisfaction driver (β = 0.42, p < 0.001)
• Statistical validation confirms robust model performance (R² = 0.68)

---

Results & Outputs

Generated Artifacts

• data/final/rss_scores.csv - RSS by station/line/time period
• reports/figures/ - Visualizations (heatmaps, time series, equity charts)
• reports/presentation.pdf - Final academic presentation

Dashboard Features

• Interactive line/station filtering
• Time-period comparisons
• Equity analysis breakouts
• Indicator contribution charts

---

Repository Structure

Visualizing-the-T/
├── data/
│   ├── raw/              # Original MBTA datasets
│   ├── processed/        # Cleaned data
│   └── final/            # RSS scores
├── notebooks/            # Analysis notebooks (01-05)
├── src/                  # Python modules
├── app/                  # Streamlit dashboard
├── reports/              # Figures and presentation
├── config/               # Weights configuration
└── requirements.txt

---

Dependencies

• Data Processing: pandas, pyspark, numpy
• Modeling: scikit-learn, scipy, statsmodels
• Visualization: matplotlib, seaborn, plotly
• Dashboard: streamlit
• Configuration: pyyaml

See requirements.txt for full list with versions.

---

Team Contributions

• Eric Fu: Data engineering, PySpark ETL pipelines, infrastructure setup
• Yuchen Cai: Feature engineering, Streamlit app development, dashboard design
• Dhanrithii Deepa: Statistical validation, equity analysis, hypothesis testing

---

License

MIT License - see LICENSE file for details

---

Acknowledgments

Data provided by MBTA Open Data Portal