THE RESULT
What the data shows
The largest computed metric is 421.3 for 1 / weather 1; the smallest is 100 for 1 / weather 4. Metric: mean_rentals (Rentals per observed hour).
Source weather categories are kept as codes. Sample sizes vary; this is a contemporaneous association, not a forecast or causal weather effect.
| workingday | weathersit | observed hours | mean rentals | median rentals | group |
|---|---|---|---|---|---|
| 0 | 1 | 956 | 228.1 | 182 | 0 / weather 1 |
| 0 | 2 | 325 | 197.9 | 152 | 0 / weather 2 |
| 0 | 3 | 104 | 120.6 | 85 | 0 / weather 3 |
| 1 | 1 | 1,873 | 421.3 | 392 | 1 / weather 1 |
| 1 | 2 | 817 | 364.2 | 329 | 1 / weather 2 |
| 1 | 3 | 292 | 230.2 | 180 | 1 / weather 3 |
| 1 | 4 | 2 | 100 | 100 | 1 / weather 4 |
THE METHOD
From source to answer
Restrict to declared commute hours and stratify by working-day status.
The Python source exposes this study’s transformations. The complete project download includes shared preparation and evaluation routines.
THE NEXT DECISION
What follows from the finding
Consider weather-aware planning only after obtaining historical forecasts available at decision time.
Where the conclusion stops
Historical system rentals measure realized use, not unmet demand or station inventory. Missing hours are unknown. Weather associations do not establish causal effects.
Related studies may reuse observations or holdouts. These are historical analyses; associations and backtests do not demonstrate commercial impact. Further model tuning needs new, untouched evaluation data.