Workplace

>70%

machine learning model predicted cognitive symptoms from office environment data with reasonable accuracy

Exact: exceeds 70%

XGBoost model accuracy for predicting cognitive symptoms generally exceeded 70%

Using about four months of continuous indoor environmental quality (IEQ) data and ecological momentary assessments (EMAs) from ten office workers, XGBoost models were built per symptom and participant. Under random train-test splitting, accuracy generally exceeded 70%, though performance declined under temporal splitting, reflecting real-world challenges of predicting future states from past observations.

Under random train-test splitting, accuracy generally exceeded 70% with F1 scores ranging from 40% to 80%.
Tao Zang et al., 2026, Indoor Environments

Machine-extracted, quote-verified. Report an error

Related findings

Read more in

Empty open-plan office with light-wood planted partitions and task chairs Workplace

What your office costs

Four design variables that move cognitive performance and who pays for them

22 May 2026 · 14 min · 25 sources
Empty classroom with wooden chair-desks and a full-height window onto trees Education

What school spaces do to children

Where the evidence on classroom air, acoustics, light and green is robust, where it is thin, and what to measure before the build.

12 May 2026 · 13 min · 18 sources

More from The Built Review

Silhouette of a person sitting at a floor-to-ceiling window with a view over Potsdamer Platz in Berlin Workplace

Germany's missing indoor-air bill

France, Britain and Australia have priced bad indoor air. Germany's missing number is a political choice, not a methodological limit.

10 Jun 2026 · 12 min · 14 sources
All reports

← All findings