Interval estimation of construction cost at completion using least squares support vector machine

Abstract

Completing a project within the planned budget is the bottom-line of construction companies. To achieve this goal, periodic cost estimation is vitally important not only in the planning phase, but also in the execution phase. Due to high uncertainty in operational environment, point estimation of project cost is oftentimes not sufficient to assist the decision-making process. This study utilizes Least Squares Support Vector Machine (LS-SVM), machine learning based interval estimation (MLIE), and Differential Evolution (DE) to establish a novel model for predicting construction project cost. LS-SVM is a supervised learning technique used for regression analysis. MLIE is employed for inference of prediction intervals. Moreover, our model deploys DE in the cross validation process to search for the optimal values of tuning parameters. The newly developed model, named as EAC-LSPIM, yields results consisting of a point estimate coupled with lower and upper prediction limits, at a certain level of confidence, to accentuate uncertainty. Simulation and performance comparison demonstrate that the new model is capable of delivering accurate and reliable forecasting results.

Keyword : construction management, prediction interval, estimate at completion, least squares support vector machine, differential evolution, machine learning