AI-supported decision framework for sustainable reconstruction: Case study • on TOKI housing after the 2023 Kahramanmaras,earthquake

Abstract

This study presents a hybrid analytical and machine learning-based framework to evaluate and classify the electricity performance of standardized TOKI center dot housing units planned for reconstruction in the aftermath of the February 6, 2023, Kahramanmaras,earthquakes. While a standardized building model was analyzed using dynamic energy simulation (DesignBuilder) for 11 affected provinces, machine learning techniques were integrated to enhance the interpretability and decision support capabilities of the output. According to local climate data and building specifications, annual electricity consumption was simulated, and units were classified into 'low' or 'high' consumption categories using thresholds defined by T & uuml;rkiye's Energy Market Regulatory Authority (EPDK). To improve classification reliability and computational efficiency, a wrapper-based feature selection approach was employed. The Whale Optimization Algorithm (WOA), guided by K-Nearest Neighbors (KNN) fitness evaluation, was used to identify a subset of the most relevant features, and a Support Vector Machine (SVM) was trained on this reduced feature set. The WOA-KNN-SVM model outperformed the baseline SVM classifier across all performance metrics, achieving 98.2 % classification accuracy, with notable improvements in sensitivity, specificity, and Matthews Correlation Coefficient. The results demonstrate that this integrated methodology can effectively support climate-sensitive and energy-efficient design decisions for mass housing in disaster-prone regions. By providing a replicable and scalable decision-support tool aligned with real-world tariff structures, the proposed approach contributes a novel perspective to post-disaster sustainable reconstruction planning.