Data and code for "Surrogate-Assisted Optimization of Parametric Facade Louvres for Annual Solar-Gain Reduction: A Grasshopper-Python Gaussian Process Framework"

This replication package contains the numerical data, source code, corrected trained Gaussian Process Regression model, external-validation evidence, benchmark comparisons, and optimization records supporting the associated research article. The controlled case examines annual incident solar-energy minimization for an east-facing parametric facade-louvre system in Dhahran, Saudi Arabia. The archive includes a 150-case Latin Hypercube Sampling training dataset, an independently seeded 30-case external-validation dataset, a documented correction record for LHS_0150, regenerated GPR predictions and uncertainty intervals, RBF-SVR and Random Forest prediction benchmarks, a 2,762-row corrected-model replay of the archived surrogate candidate sequence, a 2,854-row direct-simulation optimization log, 637 paired unique geometries, final direct verification, portable Python scripts, model metadata, environment files, and a documented Grasshopper-Ladybug workflow. The deposited model achieves external R2 = 0.999575, RMSE = 44.07 kWh, MAE = 35.68 kWh, and MAPE = 0.523%. Both optimization routes retain the selected design of 0.90 m and -45 degrees. Direct verification gives 2,153 kWh compared with the corrected GPR prediction of 2,018.75 kWh. The README documents file relationships, variables, units, random seeds, reproduction steps, correction provenance, and interpretation limits.