Cost estimation in residential building project based on LSSVM with PCA

Adequate estimation of construction cost is a vital task for throughout construction projects especially in the pre-planning stage considered by the investors, managers, even owners. Residential building project accounts for a high proportions of construction projects in most countries. As a result, its cost estimation has attracted a great deal of attention. The completed projects contain a lot of useful information for reference to the cost estimation. The purpose of this study intends to propose and validate a prediction model for cost performance of residential construction projects using the information from completed projects. The model which is based on the least squares support vector machine (LS-SVM) with principal component analysis (PCA) has been developed and validated by 290 sets of data from surveying for 25 different construction companies. 21 influencing factors related to the cost of the construction are identified and selected, whose correlation analysis is performed to determine PCA reduction dimensionality method needs to be adopted. The MAE, RMSE, MAPE and R2 are introduced to evaluate the performance of the prediction model. The experimental results show that the proposed model allows estimating the cost of residential projects with satisfactory performance.