Code and data for: Resilient Modulus of SCBA-Modified Granular Subbases: Experimental and Random Forest Analysis

Reproducible Python pipeline (Jupyter notebook) and experimental dataset (900 cyclic-triaxial records: 3 specimens at 0, 5 and 10 % sugarcane bagasse ash x 15 loading sequences x 20 readings) for the prediction of the axial resilient modulus of a granular subbase.

The pipeline excludes all variables derived from the resilient cycle (target leakage), screens 25 invalid readings, and compares a Random Forest with the classical k-θ law under four validation schemes: random split (upper bound), block hold-out, leave-one-sequence-out (primary) and leave-one-SCBA-level-out. Code: MIT license. Data: CC BY 4.0.