Pre-blast joint descriptors and median fragment size: analysis data and validation outputs

Published: 1 November 2026| Version 1 | DOI: 10.17632/bk2gvyvdb6.1
Contributor:
yh c

Description

This dataset accompanies the study of the incremental predictive value of pre-blast joint descriptors for median fragment size. It contains 71 analysis records representing 17 engineering configurations. Each record includes 14 conventional engineering variables, 10 two-dimensional joint descriptors and median fragment size measured using Split-Desktop. The files provide the analysis table, variable definitions, configuration assignments, saved out-of-fold predictions, validation test-fold membership, configuration-specific performance, 2000 paired bootstrap draws and 6000 joint-block perturbation evaluations. Predictions cover repeated record-level nested cross-validation, leave-one-configuration-out validation and the processing-sensitivity analyses reported in the manuscript. Fragment sizes and saved prediction errors are expressed in metres; the manuscript reports prediction errors in millimetres. File contents, units and validation tags are documented in README.md and Variable_Definitions.csv.

Files

Steps to reproduce

Read Analysis_Data.csv together with Variable_Definitions.csv. Match records to engineering configurations using record_id in Engineering_Configurations.csv. Use Out_of_Fold_Predictions.csv and Validation_Test_Folds.csv to identify the model, feature set, validation condition, repeat and test fold. Calculate prediction errors using the weighting and aggregation definitions given in the manuscript. Fragment sizes and saved error values are in metres; multiply by 1000 to obtain millimetres. Configuration_Performance.csv, Bootstrap_Draws.csv and Joint_Block_Perturbations.csv provide the saved outputs underlying the configuration comparisons, bootstrap summaries and joint-block perturbation results.

Institutions

Categories

Machine Learning, Rock Mechanics, Mining Engineering

Licence