| Title: | Seven Data Sets for Urban and Built-Environment Statistics |
|---|---|
| Description: | Seven documented data sets from transport, traffic safety, urban planning, construction and architectural engineering. The package provides fixed, redistributable snapshots with consistent variable names. Each help page records the source, licence, unit of observation, transformations and limitations of its data set. Sources include Yeh (2018) <doi:10.24432/C5J30W>, Tsanas and Xifara (2012) <doi:10.24432/C51307>, Yeh (1998) <doi:10.24432/C5PK67>, Seoul Bike Sharing Demand (2020) <doi:10.24432/C5F62R>, and Singh and Chaudhari (2018) <doi:10.24432/C5P605>. |
| Authors: | Mario Cools [aut, cre, cph] (ORCID: <https://orcid.org/0000-0003-3098-2693>) |
| Maintainer: | Mario Cools <[email protected]> |
| License: | MIT + file LICENSE |
| Version: | 0.1.0 |
| Built: | 2026-10-10 12:52:16 UTC |
| Source: | https://github.com/mariocools/urbstatdata |
Provides seven fixed, documented data sets. Each data-set help page records the unit of observation, variables, source, licence, transformations and limitations. The package exports no functions.
Grouped victim counts by exact combinations of accident date, time, road, victim and administrative characteristics. Codes were preserved, French/Dutch labels retained, and English variable names added.
be_accidentsbe_accidents
A data frame with 82,876 rows and 45 variables. It includes:
Accident date and hour.
Day code and French/Dutch labels.
Total victim count in the grouped cell.
Outcome counts.
Derived count: seriously injured plus dead within 30 days.
Derived count: uninjured plus slightly injured.
Built-up-area code and labels.
Victim-type code and labels.
Road-user code and labels.
Road-type code and labels.
Light-condition code and labels.
Age-group code and labels.
Municipality code and labels.
Administrative-district code and labels.
Province code and labels.
Region code and labels.
Sex code and labels.
A row is a grouped table cell, not necessarily one victim or one
crash. Consequently, binary individual-level logistic regression is not
appropriate. The data support contingency tables and grouped-binomial or
count analyses when frequencies are used correctly. Outcome components sum to
victims_total. Administrative codes are character strings to retain
leading zeroes.
Statbel and Belgian Federal Police, 2022 road-accident victim table, licensed CC BY 4.0. Stable source snapshot retrieved from the rWSBIM1207 R package at Git commit 3561033e7a99182664a0a20f603382b5539941f0, https://github.com/UCLouvain-CBIO/rWSBIM1207/tree/3561033e7a99182664a0a20f603382b5539941f0.
Counts for nine cadastral building and dwelling characteristics, four building types, Belgium as a whole and the three regions. English labels were added while the French source labels were retained.
be_building_stockbe_building_stock
A data frame with 144 rows and 10 variables:
Stable English code for the reported characteristic.
Original French characteristic label.
English and source-language country labels.
English and source-language region labels; blank source regions were labelled as the Belgium total.
Reference year (2025).
English and source-language building-type labels.
Reported count; missing where the source table does not provide a cell.
Rows are aggregate table cells, not individual buildings. Analyses must avoid individual-level or causal interpretations. Missing counts are source-table omissions, primarily footprint categories for apartment buildings, and must not be recoded as zero.
Statbel, Cadastral statistics of the building stock, official be.STAT API view 43d7cdce-3647-4f5c-86f1-a4e0c864f692, retrieved 10 August 2026. Licensed CC BY 4.0, https://bestat.statbel.fgov.be/bestat/api/views/43d7cdce-3647-4f5c-86f1-a4e0c864f692/result/CSV.
Ecotect simulations for twelve residential-building forms under different orientations and glazing arrangements. Variable names and units were standardised; values were retained.
building_energybuilding_energy
A data frame with 768 rows and 10 variables:
Dimensionless relative compactness.
Areas in square metres.
Overall height in metres.
Coded orientation (2 through 5).
Glazed-area fraction.
Coded glazing-area distribution (0 through 5).
Simulated loads in kWh per square metre.
A row is a simulated design configuration, not an observed occupied building. Results should not be generalised beyond the simulation design and assumptions. Treat the orientation and glazing-distribution codes as factors when comparing their levels.
Tsanas, A. and Xifara, A. (2012), Energy Efficiency, UCI Machine Learning Repository, doi:10.24432/C51307. Licensed CC BY 4.0.
Concrete-mixture composition, test age and measured compressive strength. Variable names and units were standardised; values were retained.
concrete_strengthconcrete_strength
A data frame with 1,030 rows and 9 variables:
Cement content in kg per cubic metre.
Blast-furnace slag in kg per cubic metre.
Fly ash in kg per cubic metre.
Water in kg per cubic metre.
Superplasticizer in kg per cubic metre.
Coarse aggregate in kg per cubic metre.
Fine aggregate in kg per cubic metre.
Age at testing, in days.
Compressive strength in megapascals.
A row is one tested mix. Relationships are strongly nonlinear, so a single untransformed linear age effect is mainly useful as a diagnostic starting point. A comparison with a benchmark is not a certification or structural-safety decision.
Yeh, I.-C. (1998), Concrete Compressive Strength, UCI Machine Learning Repository, doi:10.24432/C5PK67. Licensed CC BY 4.0.
Time-ordered measurements from a controlled room-sensor experiment, paired with the observed number of occupants.
room_occupancyroom_occupancy
A data frame with 10,129 rows and 20 variables:
Measurement date (ISO string) and clock time.
Four temperature-sensor readings in degrees Celsius.
Four light-sensor readings.
Four sound-sensor readings.
CO2 concentration in parts per million.
Estimated CO2 slope.
Passive-infrared sensor indicators.
Observed number of room occupants, 0 through 3.
Derived indicator: 1 when occupancy_count > 0, otherwise 0.
A row is a measurement time, not an independent sampled room. The
derived occupied field is provided as a convenient binary outcome.
Serial dependence, sensor placement and the controlled setting limit ordinary
independent-observation inference and external generalisation.
Singh, A. P. and Chaudhari, S. (2018), Room Occupancy Estimation, UCI Machine Learning Repository, doi:10.24432/C5P605. Licensed CC BY 4.0.
One year of hourly public bicycle-rental demand in Seoul with weather and calendar predictors. Dates were converted to ISO strings and variable names and units were standardised.
seoul_bikesseoul_bikes
A data frame with 8,760 rows and 14 variables:
Calendar date as an ISO YYYY-MM-DD string.
Number of bicycles rented during the hour.
Hour of day, 0 through 23.
Temperatures in degrees Celsius.
Relative humidity in percent.
Wind speed in metres per second.
Visibility measured in 10-metre units.
Solar radiation in MJ per square metre.
Rainfall in millimetres and snowfall in centimetres.
Season label.
Holiday-status label.
Whether the rental system was functioning.
Rows form a time series and are not independent random observations. Non-functioning periods can generate structural zeroes. Count analyses must therefore consider overdispersion, temporal dependence and service status.
Seoul Bike Sharing Demand (2020), UCI Machine Learning Repository, doi:10.24432/C5F62R. Licensed CC BY 4.0.
Market-historical records collected from the Sindian District of New Taipei City, Taiwan. Variable names were standardised and units were made explicit; values were otherwise retained.
taipei_housingtaipei_housing
A data frame with 414 rows and 8 variables:
Source row identifier.
Transaction year expressed as a decimal year.
Age of the house, in years.
Distance to the nearest MRT station, in metres.
Number of convenience stores in the living circle.
WGS84 coordinates.
Unit-area price in 10,000 New Taiwan dollars per ping.
A row is a transaction. The observational data support descriptive, associational and predictive analyses, but do not identify causal effects of MRT proximity or nearby services. A ping is approximately 3.306 square metres.
Yeh, I.-C. (2018), Real Estate Valuation, UCI Machine Learning Repository, doi:10.24432/C5J30W. Licensed CC BY 4.0.