Construction Cost Prediction Model for University New Campus Projects and Multi-Method Comparative Study—A Case Study of a University New Campus Expansion Project in Qingdao

Abstract

The new campus construction projects of universities have large investment scales and long construction periods, and the accuracy of cost estimation directly affects the arrangement of funds and the effectiveness of cost control. This paper takes a new campus construction project of a university in Qingdao as the research object, collects complete data of 13 individual buildings, and respectively builds four cost estimation models: multiple linear regression (MLR), support vector machine (SVM), random forest (RF), and BP neural network (BPNN). The performance of these models is compared through 10-fold cross-validation. The research results show: (1) The random forest model has the highest prediction accuracy, with R²=0.892, RMSE=128.6 yuan/m², MAPE=3.24%, significantly superior to the other three models; (2) The building area and floor count are the key factors affecting the cost, with a cumulative contribution rate of 71.2%; (3) Compared with traditional linear regression, the prediction errors of these three machine learning methods are on average reduced by 28.6%. The constructed prediction model in this paper can provide decision support for cost control of university infrastructure projects and also offer practical references for cost management of similar projects.