Title: Predicting cost overrun in building construction projects using support vector machine
Authors: Oluwatomisin Ayomiposi OYELEYE , Olumuyiwa Samson ADERINOLA
Volume: 10
Issue: 6
Pages: 24-33
Publication Date: 2026/06/28
Abstract:
This study looks at how well support vector tools forecast cost overruns in building projects using information from historical records. It examines the efficacy of Support Vector Machines (SVM) in forecasting construction project cost overruns by means of a measurable approach in getting its data. The independent variables of number of stories, entire floor area(m2), scope changes, provisional sum, tendering type, and initial contract sum were extracted from historical records. SVM model was compared using Decision Tree and Random Forest. The results revealed that the SVM model produced accurate construction cost predictions with 0.071, 0.099, and 0.693 of MSE, RMSE and R2 respectively on the accuracy test data. The lowest MSE and RMSE values indicate that SVR had the smallest prediction error, and the highest Rē score of 0.693 which means that SVR explained approximately 69.3% of the variance in cost overruns. This superior performance suggests that SVR was particularly effective at modeling the data and capturing the underlying trends and relationships. When considered collectively, inaccurate initial cost estimation, Material price fluctuations, Storey Count, and Changes in scope are reliable indicators of cost overruns in the construction sector. The resulted SVM model in this work can be applied as yard stick to predict potential cost overruns for Nigerian construction projects.