Cement 28-Day Compressive Strength Prediction Results

Statistical results from a study proposing a Transformer-XGBoost model for predicting 28-day cement compressive strength. The method was validated using real-world strength testing data from cement plants, achieving an average R² of 0.94 in Monte Carlo cross-validation.

Use Cases

• Validate the performance of the proposed Transformer-XGBoost model using the reported R² and RMSE metrics.
• Analyze the impact of Transformer feature extraction on model accuracy by comparing the 5.62% R² increase and 22.33% RMSE decrease.
• Benchmark small-sample prediction methods using the reported 5-fold cross-validation results where XGBoost achieved an average R² of 0.93.

Strengths

• Model performance is quantified with specific metrics, including a 5.62% increase in R² and a 22.33% decrease in RMSE.
• Validation involved 5-fold cross-validation and 25 Monte Carlo cross-validation runs, providing robust performance estimates.
• The dataset is openly shared under a CC BY 4.0 license, facilitating reuse and verification.

Limitations

• The dataset is extremely small at 9.5 KB, indicating it contains only summary statistical results, not the raw experimental data.
• The specific number of rows, columns, and features used in the underlying model training is unknown.
• The geographic origin of the real-world cement plant data is not specified, limiting assessment of regional applicability.

Provenance

Source

Dianyuan Ju via figshare.

Collection Method

Results from a proposed machine learning method validated on real-world 28-day cement strength testing data.

Freshness

Last updated March 24, 2026.