A forecasting algorithm predicts future values from time-series data. Before applying one to road-traffic measurements, establish what each row represents, how records vary across locations and time, and whether observations arrive at consistent intervals. section 8.3
Prepare the data by checking:
- Granularity, including whether each row represents a location-day or another unit. section 8.3
- Geographic and temporal coverage, including unequal record lengths across locations. section 8.3
- Gaps and inconsistent intervals, which can prevent many forecasting algorithms from working reliably. section 8.3
- Traffic distributions, daily and weekly variation, yearly seasonality, and growth trends before narrowing analysis to better-covered locations. section 8.3
Road-traffic forecasting
Bicycle counts from multiple locations can be explored with line charts to reveal temporal behavior, expose data limitations, and support conversations about additional datasets and priorities. section 8.3
