A Neural Network Model Based on Support Vector Machine for Conceptual Cost Estimation in Construction Projects

Subject Areas : Design of Experiment

Behnam Vahdani ¹ , Seyed Meysam Mousavi ² , Morteza Mousakhani ³ , Mani Sharifi ⁴ , Hassan Hashemi ⁵

1 - Instructor, Industrial engineering research center, Qazvin Branch, Islamic Azad University, Qazvin, Iran
2 - Ph.D. Student, Young Researches Club, South Tehran Branch , Islamic Azad University, Tehran, Iran
3 - Associate Professor, Department Management and Economics, Science and Research Branch, Islamic Azad University, Tehran, Iran
4 - Assistant Professor, Faculty of Industrial and Mechanical Engineering, Qazvin Branch, Islamic Azad University, Qazvin, Iran
5 - M.Sc, Young Researches Club, South Tehran Branch, Islamic Azad University, Tehran, Iran

Received: 2011-08-12 Accepted : 2011-12-22 Published : 2012-03-15

Keywords: Construction Projects, Conceptual cost estimation, Support vector machine, Cross validation,

Abstract :

Estimation of the conceptual costs in construction projects can be regarded as an important issue in feasibility studies. This estimation has amajor impact on the success of construction projects. Indeed, this estimation supports the required information that can be employed in costmanagement and budgeting of these projects. The purpose of this paper is to introduce an intelligent model to improve the conceptual costaccuracy during the early phases of the life cycle of projects in construction industry. A computationally efficient model, namely supportvector machine model, is developed to estimate the conceptual costs of construction projects. The proposed neural network model is trainedby a cross validation technique in order to produce the reliable estimations. To demonstrate the performance of the proposed model, twopowerful intelligent techniques, namely nonlinear regression and back-propagation neural networks (BPNNs), are provided. Their resultsare compared on the basis of the available dataset from the related literature in construction industry. The computational results illustratethat the presented intelligent model performs better than the other two powerful techniques.