Urban Transit Lab
Forecast Comparison Research Roadmap Weather GitHub
Demand Intelligence

Forecast micromobility demand

Use GBFS historical data, weather patterns, and event signals to predict bike and scooter demand with confidence. Optimize fleet deployment and station rebalancing in real time.

Open Forecast Learn More
What We Forecast

Accurate, Actionable Predictions

Trip Volume

24-hour demand patterns segmented by hour, day type, and zone. Peak demand windows with 92%+ confidence on short-term forecasts.

Station Availability

Predict bike/scooter distribution across station networks. Identify rebalancing needs before stockouts occur and optimize dispatch.

Geographic Hotspots

Zone-level demand forecasts. Focus resources on high-demand areas: downtown, transit hubs, university districts, and residential corridors.

Confidence Intervals

Every forecast includes estimated uncertainty. Weather events, one-off incidents, and edge cases are surfaced explicitly.

Inputs & Signals

GBFS Trip Data 90+ days of historical bike/scooter usage from live feeds
Weather Patterns Temperature, precipitation, wind speed effects on demand
Day-Type Effects Weekday/weekend/holiday demand splits, seasonal curves
Event Signals Recurring events (commute peaks, sports games, festivals)
Geographic Context Zone demand patterns, station proximity, land use type
Time Series Intraday patterns, day-to-day volatility, trend direction
Project

What Can Be Accurately Forecast

Micromobility demand is highly predictable in the short term (hours to days ahead) when you have the right data. GBFS trip history shows clear, repeating patterns: commute peaks, weekend leisure demand, weather sensitivity, and stable day-of-week effects all drive usage in measurable ways.

Strong signals: Historical patterns dominate. If Tuesday at 8 AM always shows peak commute demand, next Tuesday at 8 AM will too—weather and calendar type accounted for. Intraday forecasts (1–7 days out) reach 15–25% MAPE (mean absolute percentage error) with baseline time-series models. Adding structured weather and event data tightens this further.

Where accuracy falls off: One-off events (stadium games, street closures) create noise outside the historical distribution. New infrastructure (bike lane, office building) shifts the baseline in ways the model can't see. Black-swan disruptions (extreme weather, service outages) require explicit handling. But for routine operations—day-to-day demand planning, fleet sizing, rebalancing schedules—GBFS data is predictive enough to optimize real decisions.

This tool bridges the gap: it shows what's reliably forecastable from public data and where external signals (event APIs, weather feeds) unlock additional accuracy. Use it to understand your demand surface and plan accordingly.