UCI BIKE SHARING — REAL-DATA WORKFLOW EXAMPLE
Downloaded: 2 November 2026

PRIMARY SOURCE AND LICENSE
Fanaee-T, H. (2013). Bike Sharing [Dataset]. UCI Machine Learning Repository.
DOI: https://doi.org/10.24432/C5W894
Dataset page: https://archive.ics.uci.edu/dataset/275/bike+sharing+dataset
Official archive: https://archive.ics.uci.edu/static/public/275/bike%2Bsharing%2Bdataset.zip
UCI states CC BY 4.0: https://creativecommons.org/licenses/by/4.0/
Associated paper: Fanaee-T, H., and Gama, J. (2013), Event labeling combining ensemble detectors and background knowledge. https://doi.org/10.1007/s13748-013-0040-3

The archive includes daily counts from Capital Bikeshare in Washington, DC, 2011–2012, with weather/calendar covariates. day.csv is used, not hour.csv. The page's headline instance count concerns the broader dataset; daily row count was computed from the actual file.
Original day.csv is preserved byte-for-byte.
SHA-256: a6bcf826782d3c0fbfdcbeead17cd0884185a0dafe8ff10cd48a874ee7ba18be

ATTRIBUTION FOR THE WEBSITE
Data: Hadi Fanaee-T / UCI Machine Learning Repository, Bike Sharing (2013), CC BY 4.0. Analysis and presentation added for this illustrative workflow. No institutional endorsement or product validation is implied.

METADATA DISCREPANCIES
The archive README describes temp as divided by 41 and gives season labels that disagree with the current UCI page. The current page describes a different temperature normalization, qualified as hourly. Therefore the example does not claim daily temperature in Celsius. It retains normalized temp, labels a precise 0.50→0.60 contrast, and derives month from dates. Wind is likewise presented in normalized units; the archive does not specify a physical unit. Relative humidity is recorded divided by 100, so 0.10 corresponds to ten percentage points.

WHAT THIS EXAMPLE DOES AND DOES NOT ESTABLISH
This is an independently reproducible statsmodels analysis of public observations. It does not run an OptiMod production engine and must not be presented as a verified OptiMod product screenshot or performance benchmark. It demonstrates a proposed workflow using real data and real calculations.
The target is recorded rental trips/day, not unique riders, unconstrained demand or required fleet size. Counts may reflect capacity and service disruptions. Both model predictions and effect estimates condition on observed weather. They are not operational forecasts using weather information available in advance.

METHOD REFERENCES
Negative-binomial NB2 model: https://www.statsmodels.org/stable/generated/statsmodels.discrete.discrete_model.NegativeBinomial.html
Heteroskedasticity/autocorrelation robust covariance: https://www.statsmodels.org/stable/generated/statsmodels.stats.sandwich_covariance.cov_hac.html
Model execution uses pinned statsmodels 0.14.6; current online stable documentation can describe a newer release.
