compressive strength

Compressive strength is the load per unit area a cement or mortar specimen carries before it fails, reported in megapascals at fixed curing ages. It is the standard performance measure in cement characterization and the usual prediction target in concrete machine-learning work.

It appears twice here. The 27-Year Multi-Producer European Cement Characterization Dataset carries 50 variables covering "chemistry, fineness, setting behaviour and 2/7/28-day strength" - 476 samples from 23 anonymized European producers, measured 1997 to 2024.

The Concrete Compressive Strength Dataset on the UCI Machine Learning Repository maps its 1,030 laboratory mix records "to compressive strength in MPa", making it the label column of a benchmark that has outlived most of its contemporaries.

Strength is a specification and a safety property, so a dataset that misstates its basis makes every downstream model untrustworthy.

• Age is part of the measurement. 2-, 7- and 28-day values are different numbers for the same cement; a source that collapses them into one column destroys the curing curve.

• Lab provenance varies. UCI holds 1,030 records from one laboratory tradition; the European set spans 23 producers over 27 years - neither transfers to plant QC without validation.

• Anonymization limits attribution. In the European dataset producer identities are deliberately anonymized and must not be re-identified, so supplier-level conclusions are out of scope.

• Licensing differs. UCI is commercial delivery terms 4.0 with a required citation to Yeh's paper; the European set is commercial delivery terms 4.0 naming its authors and DOI.

1. Check the curing ages are explicit. The European dataset reports 2-, 7- and 28-day strength per sample; a source without stated ages cannot be used for specification work.

1. Confirm the units. Values should be in MPa - the UCI record maps its mixes "to compressive strength in MPa".

1. Count the rows and variables. 476 samples x 50 variables (European, 1997-2024) and 1,030 instances x 9 attributes (UCI) are the documented sizes.

1. Ask which variable is the target. In the UCI benchmark strength is the label; a modeling extract that shuffles columns changes the task.

1. Verify licence and citation terms. Both datasets are commercial delivery terms 4.0; the UCI readme additionally requires retaining Prof. I-Cheng Yeh's copyright notice and citing his modeling paper.

Where to see it in context: construction materials data.

These related entries and hubs sit next to this term: - concrete mix design - construction materials data