Dataset Reviews & Methodology
Make provenance, geography, labels, missingness, and evaluation methods visible before interpreting a dataset.
Read beyond the headline.
Traceable data, reproducible evaluation, and a clear boundary between record reconciliation and substantive validation.
Questions we bring to the evidence
- What population, period, and geography do the observations cover?
- Are labels independently observed or inferred from the same signals used to predict them?
- Were evaluation data separated by time, room, or building, and were raw records preserved?
When Smart Buildings Get It Wrong: Environmental Sensors, Occupancy, and HVAC Intelligence.
Follow the evidence from sensor readings through occupancy estimates to decisions about ventilation and building operation.
Read with sources & limitationsA repeatable review framework
- Define the unit. Distinguish sensor samples, occupied intervals, source records, properties, buildings, and roads. One is not a substitute for another.
- Preserve provenance. Record source, release, field definitions, access conditions, transformations, and known omissions. Retain original labels and observations.
- Test the connection. Reconcile counts and joins, then separately evaluate label accuracy, geography, measurement quality, and the suitability of the proposed use.
- Evaluate honestly. State the comparison, temporal and geographic split, missing-data handling, outcome metrics, uncertainty, and limits to transfer.
- Publish the boundary. Separate verified findings, proposed interpretations, unresolved cases, and untested claims. Reconciliation checks alone do not certify every grouping.
For NDS property intelligence, address-linked context can inform territory research. It does not directly measure current occupancy, equipment condition, airflow, or indoor air quality. Property-to-street join verification and county recovery remain separate validation tasks; this department does not certify those results.
Sources & research records
Primary-source records accessed November 18, 2026. Access level is stated for each source.
- Luis Candanedo (2016). Occupancy Detection UCI Machine Learning Repository · DOI 10.24432/C5X01N. Dataset · repository documentation reviewed.
A binary occupancy benchmark with environmental measurements and photograph-derived reference labels. No model was retrained for this review.
DOI record ↗ - Andrew K. Persily (2022). Evaluating Ventilation Performance NIST publication record · Handbook of Indoor Air Quality. Technical book chapter · institutional abstract reviewed.
Frames ventilation evaluation as a measurement problem that varies with weather, operation, and the purpose of the assessment.
DOI record ↗