Skip to content
Contact UsStart Building

Construction & Materials

Concrete strength

Estimate concrete strength to prioritize candidate mixes for laboratory testing.
AI-generated illustration of a cylindrical concrete specimen in a compression-testing rig.
AI-generated illustration. A representative setting, not a customer or benchmark facility.
From known examples to a prediction

Available inputs

  • Seven ingredient quantities in kg/m³
  • Curing age in days

Known examples

1,030 recorded examples

Prediction target

Compressive strength in MPa
6.9%lower prediction error

vs. CatBoost (default), the strongest conventional result in this benchmark.

Concrete strengthRMSE · Lower is better

1,030 examples · 8 inputs · 30 evaluation splits

010
  1. Causilo
    3.9128
  2. CatBoost
    4.2017
  3. LightGBM
    4.2117
  4. XGBoost
    4.2222
  5. Extra trees
    5.0033
  6. Random forest
    5.1063
  7. K-nearest neighbors
    6.5484
  8. Linear regression
    8.1534

The task in practice

The public study contains 1,030 measured examples and eight inputs: cement, blast-furnace slag, fly ash, water, superplasticizer, coarse aggregate, fine aggregate and age. Each row represents a mix at a particular age.

  1. Use past measurements as context. Pair the mix ingredients and age with their measured compressive strength.
  2. Describe a candidate mix. Provide the same eight measurements for a new recipe, leaving its strength unknown.
  3. Prioritize the next physical test. Use the estimate to inform a shortlist, then verify the selected mix in the laboratory.

Scope. Validation should include the ingredients and manufacturing conditions of the intended application.

Dataset, sources and evaluation

I-Cheng Yeh, Concrete Compressive Strength (1998), UCI Machine Learning Repository. Original dataset source · CC BY 4.0.

The TabArena-curated OpenML version (46917) supplies the dataset and recorded target, ConcreteCompressiveStrength. Comparison scores come from the retained results_per_split_official.csv evaluation snapshot. These pages describe the task and compare the recorded means; source records are not republished here.

Lower RMSE means the strength estimates were closer to the held-out measurements on this dataset. Laboratory testing would establish whether those estimates improve a recipe shortlist.

Causilo uses its default configuration. Each conventional family uses its strongest reported mean across default, tuned and tuned-ensemble settings. The chart compares reported means, without a claim of statistical significance, customer deployment or operational savings.

Relative error change = (strongest conventional mean − Causilo mean) ÷ strongest conventional mean × 100. A negative change is shown as higher error.

How to read these comparisons · All configurations and original values.