Akademik Veri Yönetim Sistemi
7th International Project and Construction Management Conference, İstanbul, Türkiye, 20 Ekim 2022, ss.480-492
To protect the occupants’ lives, it is important to ensure the structural safety of the existing building stocks. For this reason, the issue of retrofitting existing buildings is a frequently encountered situation today. However, decision-making about building retrofitting or complete demolishment is a cost- and time-consuming challenge. Therefore, this study aims to identify and measure parameters that influence the decision-making process about retrofitting a building. To address this issue, first Pearson correlation was used for establishing the relationship amongst the retrofitting feature, then three ensemble machine learning models (MLs), named Random Forest (RF), Gradient Boosting (GB), and Adaboosting (AB), employed for determining the building health condition. To establish the model's effectiveness, this research collected health monitoring reports obtained from 111 school buildings within eight cities in Turkey. The important building features were used to classify the health condition of the building. Based on the results the AB outperformed both RF and GB considering accuracy, F1 score, precision, and recall metrics (79%, 79%, 80%, and 79%). Also, feature selection using Pearson correlation could improve the performance of RF and GB by up to 3%.