Forecasting Construction Price Index using Artificial Intelligence Models: Support Vector Machines and Radial Basis Function Neural Network

Abstract

Estimation of Construction Price Index (CPI) is important for a market economy and it is a measure to manage construction investment costs. This is a tool to help organizations and individuals to reduce the effort and management of expenses for construction projects by reducing time of procedures for calculating and adjusting the total investment for the estimation and evaluation of contract price. The CPI is an indicator that reflects the level of construction price fluctuations of the type of work over time. In this study, the CPI data of Son La province, Vietnam from January 2016 to March 2022 (75 dataset) has been used for the modelling. Two Artificial Intelligence (AI)  models namely Support Vector Machine (SVM) and Radial Basis Function Neural Network  (RBFN) were proposed to predict the CPI based on limited input data. Performance of the models in correctly predicting CPI was evaluated using standard statistical indicators such as Coefficient of Determination (R²), Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) based on the historical CPI data. The results show that performance of both the models is good in predicting CPI, but performance of the  SVM model (R² train = 0.915, R² test = 0.811) is the best in comparison to RBFN model (R² train = 0.985, R² test = 0.733). The proposed AI models can be used to quickly and accurately predict the CPI of an area to help management agencies, investors, construction contractors for assessing cost of construction for the purchase and development of properties/ infrastructures.