Improving classification accuracy of project dispute resolution using hybrid artificial intelligence and support vector machine models Jui-Sheng Chou ⇑ , Min-Yuan Cheng, Yu-Wei Wu Department of Construction Engineering, National Taiwan University of Science and Technology, 43, Sec. 4, Keelung Rd., Taipei 106, Taiwan article info Keywords: Classification Hybrid intelligence Support vector machines Fuzzy logic Genetic algorithm Dispute resolutions Construction management abstract Support vector machines (SVMs) have been applied successfully to construction knowledge domains. However, SVMs, as a baseline model, still have a potential improvement space by integrating hybrid intelligence. This work compares the performance of various classification models using the combination of fuzzy logic, a fast and messy genetic algorithm, and SVMs. A set of public–private partnership projects was collected as a real case study in construction management. The data were split into mutually inde- pendent folds for cross validation. Experimental results demonstrate that the proposed hybrid artificial intelligence system has the best and most reliable classification accuracy at 77.04%, a 24.76% improve- ment compared with that of SVMs in predicting project dispute resolution (PDR) outcomes (i.e., media- tion, arbitration, litigation, negotiation, and administrative appeals) when the dispute category and phase in which a dispute occurs are known during project execution. This work demonstrates the improvement capability of hybrid intelligence in classifying PDR predictions related to public infrastructure projects. Ó 2012 Elsevier Ltd. All rights reserved. 1. Introduction Construction projects, by their nature, are highly uncertain and intrinsically experience-oriented. Experience as tacit knowledge should be preserved and managed properly. Artificial intelligence (AI) has the ability to simulate human inference and capture expe- rience via state-of-the-art analytical tools. Artificial intelligence re- fers to computing technologies that handle complex or poorly structured problems using such tools as Artificial Neural Networks (ANNs) and Support Vector Machines (SVMs). As AI-based models can cope with tasks at which humans excel, utilizing an AI infer- ence model is a viable alternative and a promising solution to the field of construction engineering and management (CEM). In the CEM knowledge domain, project dispute resolutions are challenging when making early strategic decisions. Although the public sector is risk-averse and typically avoids financial guaran- tees, government support for construction projects is common in both developing and developed countries. However, due to the high risks associated with the construction industry, project de- lays, budget overruns, poor construction quality, and legal issues during implementation, construction, operating, and transfer phases can cause project disputes among stakeholders. Many stud- ies (Arditi & Pulket, 2010; Chen, 2008; Chen & Hsu, 2007; Cheng, Tsai, & Chiu, 2009; Chou, 2012) have demonstrated that an effi- cient, effective, and fair warning model of potential disputes or construction claims during the early planning phase is essential to project success. Research has shown that ANNs and SVMs are effective tools for solving construction management (CM) problems (Arditi & Tokd- emir, 1999a; Chen & Hsu, 2007; Chou, 2012; Lin & Hsu, 2002). However, ANNs have difficulties in identifying the optimal archi- tecture, number of hidden layers, number of neurons in layers, and learning rate (Chou, 2012). Additionally, their training process is typically optimized locally and time consuming. This work at- tempts to improve the prediction accuracy of SVMs when applied for construction dispute resolution. Despite the superiority of SVMs, no study has investigated accuracy improvements by fusing hybrid intelligence (i.e., a fast and messy genetic algorithm (fmGA) and fuzzy logic (FL) in this work) with SVMs, the baseline model. Specifically, FL, a popular AI technique invented by Zadeh in the 1960s, has been used in forecasting, decision-making, and action control in environments characterized by uncertainty, vagueness, presumptions, and subjectivity (Bojadziev & Bojadziev, 2007). Fuz- zy logic consists of a set of rules that relates a set of inputs to a set of outputs. Quantitative relationships are established through membership functions (MFs) between actual variable values and qualitative, linguistic variables used in ‘‘if-then’’ rules (Zadeh, 1973). The goal of an FL paradigm is to mimic the human inference. The fmGA-based approach developed by Goldberg, Deb, Kar- gupta, and Harik (1993) is known for its flexibility in allowing hybridization with other methodologies to obtain enhanced solu- tions. The primary difference between an fmGA and other genetic 0957-4174/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.eswa.2012.10.036 ⇑ Corresponding author. Tel.: +886 2 2737 6321; fax: +886 2 2737 6606. E-mail address: jschou@mail.ntust.edu.tw (J.-S. Chou). Expert Systems with Applications 40 (2013) 2263–2274 Contents lists available at SciVerse ScienceDirect Expert Systems with Applications journal homepage: www.elsevier.com/locate/eswa