Akademik Veri Yönetim Sistemi
DÜMF Mühendislik Dergisi, cilt.15, sa.3, ss.755-765, 2024 (Hakemli Dergi)
Accurately estimating the construction contract price is a necessary step for correctly determining project budgets and ensuring efficient use of resources. In this study, contract price in public construction tenders are estimated using structural project variables. The variables applied in the study are created by adding the quantities of columns, shear walls, and beams to variables commonly used in the literature for cost estimations. Six different machine learning algorithms are employed as machine learning algorithms. These are Support Vector Machine (SVM), Decision Tree (DT), eXtreme Gradient Boosting (XGBoost), k-Nearest Neighbors (KNN), Artificial Neural Network (ANN), and Random Forest (RF). Preprocessing methods and a series of hyperparameter optimizations are applied to enhance the predictive capability on datasets. These processes and the applied algorithms are evaluated with five different performance metrics. The Support Vector Machine (SVM) algorithm produced the best results, achieving a coefficient of determination (𝑅 2 ) of 0.8966, a Mean Absolute Percentage Error (MAPE) of 23.70, a Nash-Sutcliffe Efficiency (NSE) of 0.8956, a Mean Absolute Error (MAE) of 0.4849, and a Root Mean Square Error (RMSE) of 0.6989. This study contributes to the literature by developing machine learning models and data analysis processes for contract price approaches.