What it is
Time-series forecasting estimates future values from past observations. City teams may forecast transit demand, electricity use, water consumption, or congestion. Useful forecasts depend on stable definitions, timely data, and a clear decision window.
How it works
- 01Choose the quantity, geographic unit, and forecast horizon.
- 02Inspect missing periods, sensor changes, calendar patterns, and unusual events.
- 03Fit a baseline or model using past data and relevant explanatory variables.
- 04Test on later time periods, report uncertainty, and compare forecasts with a simple baseline.
- 01
Choose the quantity, geographic unit, and forecast horizon.
- 02
Inspect missing periods, sensor changes, calendar patterns, and unusual events.
- 03
Fit a baseline or model using past data and relevant explanatory variables.
- 04
Test on later time periods, report uncertainty, and compare forecasts with a simple baseline.
Why it matters
Forecasts help agencies prepare staff, capacity, and resources before demand arrives. Their usefulness is measured by the decisions they improve, not by a chart alone; uncertainty should be visible to planners and operators.
Typical Smart City applications
- Transit ridership and vehicle demand
- Building or district energy loads
- Congestion and travel-time ranges
- Water demand and peak-use planning
Example data inputs
Strengths
- Makes recurring patterns and changing demand easier to compare
- Supports advance staffing and capacity decisions
- Can provide ranges and scenarios rather than a single point
Limitations and risks
- Past patterns can fail after policy, service, or climate changes.
- Aggregated accuracy can hide poor forecasts for specific places or groups.
- Bad sensor data and unrecorded disruptions can create misleading certainty.
Illustrative workflow
Applied city example
A transit agency forecasts boardings by route for the next week. It compares the model with a seasonal baseline, includes scheduled service changes, and shows a range so depot teams can see when extra capacity may be needed.
Related methods
- Regression
- Demand forecasting
- Anomaly detection
- Decision support systems
Frequently asked questions
How far ahead should a city forecast?
Use the horizon that matches the decision. A short staffing forecast and a long capital plan need different data and validation.
Should the system return one exact number?
Often no. Prediction intervals or scenarios communicate uncertainty more honestly.
What is a useful baseline?
Simple seasonal averages or last-period values can reveal whether a more complex model adds practical value.