Ml codes
Description
This repository contains the machine-learning codes developed for the paper *Design parameter sensitivity and machine-learning prediction of embodied carbon and material cost intensity in early-stage commercial building design*. The scripts implement data preprocessing, feature scaling, target transformation, model training, model evaluation, and visual validation for predicting embodied carbon intensity and material cost intensity from early-stage commercial building design variables. The workflow evaluates six regression models: Artificial Neural Network (ANN), Gaussian Process Regression (GPR), Support Vector Regression (SVR), Random Forest (RF), K-nearest Neighbours (KNN), and Light Gradient Boosting Machine (LightGBM). Model performance is assessed using R², RMSE, MAE, and WAPE on a held-out testing dataset. The repository also includes code for generating predicted-versus-actual parity plots used to compare model accuracy and generalisation behaviour. The shared codes are intended to support transparency, reproducibility, and future development of early-stage surrogate modelling workflows for embodied carbon and material cost intensity prediction in commercial building design. The underlying dataset is not publicly included in this repository and is available from the corresponding author upon reasonable request.