DanielRegaladoCardoso/miami-transit-equity

Miami-Dade Transit Equity Simulator

Interactive simulation of how transit-service changes affect equity outcomes across 504 Miami-Dade census tracts.

Built for the University of Miami MSBA Capstone: AI for Equitable Public Transportation (Deloitte partnership).

What you can do

• —Adjust 4 transit-service levers (peak AM frequency, early-AM frequency, weekend/weekday ratio, rail modal share) and see per-tract equity impact.
• —Scope interventions to All / Critical-only / High+Critical / Fragile / custom GEOIDs.
• —Click any tract for a detailed narrative including demographics, service levels, and a route-level diagnostic (which bus routes serve the tract, their AM-peak headway, and the bottleneck route).
• —Before/after choropleth maps + 4×4 tier-shift matrix for each scenario.

Under the hood

• —Simulator: Luna's Sprint 3.2 engine (proportional-change formula: ratio = new_deficit / baseline_deficit, applied to Sprint 2a composite).
• —Model: XGBoost v3 predicting service_deficit from 13 features.
• —Baseline: 504 tracts × 29 cols (inner join of features + deficit + need).
• —City2Graph: GTFS-derived per-tract route-level diagnostic (Miami-Dade GTFS 2026-03-24, weekday AM peak 06:00–09:00, 500m walking buffer).

Reference scenarios (S1–S5)

Top problem routes (Critical+High tracts)

From the City2Graph diagnostic:

Credits

• —Luna (lunaasage): Sprint 3.1 baseline + Sprint 3.2 simulator
• —Lina Graf: Gradio V2/V3 dashboard
• —Daniel Regalado Cardoso: Sprint 1/2 modeling, Sprint 3.3 validation, City2Graph