Estimate at completion for construction projects Using Evolutionary Gaussian Process Inference Model

Abstract

The Estimate at Completion (EAC), which is the manager's projection of a project's total cost at its completion, is an important tool in monitoring a project's performance and risk. Executives usually make the high-level decisions on a project but they may have gaps in technical knowledge which may cause error in their decisions. This study employed an EAC estimation model to extract historical knowledge from previous construction projects. References and historical data relating to construction project costs are collected and inputted into the EAC model for it to learn from. This is then combined with recently developed AI techniques to simulate human decision-making behavior in solving management problems. This study combined a new machine learning technology based on statistic principles, Gaussian Process (GP), and Particle Swarm Optimization (PSO) to construct an Evolutionary Gaussian Process Inference Model (EGPIM). The model uses GP to determine the relationship between the input and output variables and uses the PSO optimization tool to optimize the hyper-parameters in the data function. The study investigated the application of EGPIM in determining the EAC and calculating the tendency of change in the forecast model monitor. The model provided a reliable trend of EAC estimates, which can aid project managers in improving the effectiveness of project cost controls. The learning results have been validated with real construction project data. The validation results show a better performance of the trained EGPIM model over other AI techniques such as the support vector machine and the Gaussian process.