ATE_Supplementary_Code
Description
This repository contains three Python scripts supporting the modelling and validation of outdoor thermal satisfaction. The study investigates whether multidimensional subjective perceptions provide incremental predictive information beyond objective microclimatic and spatial predictors and evaluates predictive performance on held-out survey days. The research data were collected through concurrent microclimatic measurements and questionnaire surveys at two study sites and six outdoor microenvironments over four summer and four winter survey days. The predictors comprise five objective microclimatic variables, two spatial descriptors, and 12 subjective perception variables. Thermal satisfaction is classified into three categories: dissatisfied, neutral, and satisfied. The three scripts cover seven-family machine-learning model screening, hyperparameter optimization and selection among five retained model families, and final XGBoost evaluation under leave-one-day-out validation. Adding subjective perceptions improved predictive performance across all seven model families in both seasons, although performance varied across seasons, survey days, and satisfaction classes. This repository contains analysis code and supporting documentation but does not include the complete analytical dataset. Data access requests are subject to applicable ethical and privacy restrictions.