Improving estimate at completion (EAC) cost of construction projects using adaptive neuro-fuzzy inference system (ANFIS)

Abstract

Earned value management (EVM) is a well-known technique for measuring project performance and progress. Owing to the EVM's attitude to simultaneously combine cost and time performance, project performance can be forecasted accurately, and this plays a vital role in the future of the projects. In the current study, the authors employed an adaptive neuro-fuzzy inference system (ANFIS) as a powerful prediction tool to forecast the completion cost of the projects considering the percentage of risk for qualitative variables and comparing it with other types of neural networks. Because the network structure is usually tuned based on the obtained results, a network optimization procedure is applied using a conventional method for estimating the cost-caused project breakdown. The results showed that ANFIS had a suitable performance (MSE = 0.0003), and based on the sensitivity analysis, the earned value is recognized as the most sensitive factor in the project. This study improves the general estimate of the completion formula by considering the uncertain conditions.