Project dispute prediction by hybrid machine learning techniques

Journal of Civil Engineering and Management

Yu-Hsin Lu¹  Jui-Sheng Chou²  Chih-Fong Tsai³ 

[1] Department of Accounting, Feng Chia University, 100, Wenhwa Rd. Seatwen, Taichung 40724, Taiwan;Department of Construction Engineering, National Taiwan University of Science and Technology, 43, Sec. 4, Keelung Rd, Taipei, 106, Taiwan (R.O.C.);Department of Information Management, National Central University, No. 300, Jhongda Rd Jhongli City, Taoyuan County, 32001, Taiwan;

关键词: machine learning;    clustering and classification;    hybrid intelligence;    public-private partnership;    project management;    dispute prediction;   

DOI  :  10.3846/13923730.2013.768544

来源: DOAJ

【 摘 要 】

This study compares several well-known machine learning techniques for public-private partnership (PPP) project dispute problems. Single and hybrid classification techniques are applied to construct models for PPP project dispute prediction. The single classification techniques utilized are multilayer perceptron (MLP) neural networks, decision trees (DTs), support vector machines, the naïve Bayes classifier, and k-nearest neighbor. Two types of hybrid learning models are developed. One combines clustering and classification techniques and the other combines multiple classification techniques. Experimental results indicate that hybrid models outperform single models in prediction accuracy, Type I and II errors, and the receiver operating characteristic curve. Additionally, the hybrid model combining multiple classification techniques perform better than that combining clustering and classification techniques. Particularly, the MLP+MLP and DT+DT models perform best and second best, achieving prediction accuracies of 97.08% and 95.77%, respectively. This study demonstrates the efficiency and effectiveness of hybrid machine learning techniques for early prediction of dispute occurrence using conceptual project information as model input. The models provide a proactive warning and decision-support information needed to select the appropriate resolution strategy before a dispute occurs.

【 授权许可】

Unknown