A Neural Network Model for Predicting Cost of Pre-Fabricated Housing

A Neural Network Model for Predicting Cost of Pre-Fabricated Housing

Mladen Vukomanović, Mirsad Kararić, Mladen Radujković
DOI: 10.4018/ijitpm.2014010102
(Individual Articles)
No Current Special Offers


Low performance of the construction industry stresses the need for improving current practices - especially in regard to cost. In this study the authors have found a critical set of variables for predicting total cost of pre-fabricated housing. A neural network model was applied on more than 30 projects. The model relies on 17 critical cost prediction variables. Verification, on 28 buildings, showed that: 85.7% of predicted values had the deviation lower 5%, while 10.7% had the deviation lower than 10%, in relation to the actual cost. After validating the model on new data the performances were as follows: 83.8% of predicted values had the deviation lower 5%, while 12.9% had the deviation lower than 10%. Thus, using this model, construction companies can influence project performance during project early phases, and acquire more competitive position on the market.
Article Preview


Since construction industry has a poor project management performance, especially in relation to achieving cost objectives (Al-Jibouri, 2003; Atkinson, 1999; Costa, Formoso, Kagioglou, & Alarcón, 2004; Roztocki & Weistroffer, 2005), construction companies have to be more accurate in predicting total cost of building, still during the tendering phase. Thus nowadays only 34% of projects make it on the time, within the budget and in accordance with the agreed scope (Standish-group, 2009; Vukomanović, Radujković, & Burcar Dunović, 2008). Because of this bad performance the construction has been proclaimed the worst, inefficient, plundering etc (Beatham, Anumba, Thorpe, & Hedges, 2004, 2005).

Neural networks (NN) are a form of artificial intelligence which tries to simulate behavior of human brain. The first model was published in 1943 by McCulloch and Pitts (McCulloch & Pitts, 1943), where the authors analyzed fundamental logic functions. Besides the founders; Hebb (1949), Lashley (1951), Minsky (1988) and many others served as the catalyst of change and as promoters of NNs. After initial fame, up to 1980s and the IT dawn, NN were developing in a slower pace, considering only smaller improvement (Kohonen, 1988; Rumelhart & McClelland, 1982; Werbos, 1994). Rapid development of IT stimulated further development of NN, as well.

Nowadays, NN can be also found in researches in area of construction management. E.g., Chua et al. (Chua, Loh, Kog, & Jaselskis, 1997) studied the influence of critical success factors on planned construction cost. Ling et al. (2004) defined a set of indicators that can predict project performance in Design & Build projects. They found strong correlation among 65 success factors and 11 criteria of project success. Odeh et al. (2002) studied a set of indicators in correlation with project time, in traditional, construction management procurement routes. Kog et al. (Kog, Chua, Loh, & Jaselskis, 1999) designed a model using NN in predicting the deviation [%] of project time. Iyer et al. (2005) and Iyer and Jha (2006) analyzed more than 1500 projects in India and found reliable set of factors for prediction of cost in traditional building projects. NN can be also found in many areas besides construction, e.g.: finance, ICT, medicine, transport etc, as well.

While competitive pressures are forcing construction contractors companies to produce numerous bids every day (Al-Jibouri, 2003; Iyer & Jha, 2005; Radujkovic, 1999), investors (sponsors) do not have any obligation to even consider them. This huge gap between realized and unrealized bids has unfortunately become a common practice, especially in pre-fabricated housing. This only indicates low ability of the companies in calculating the final cost. Therefore we sought to find a solution to this problem using NN approach.

Complete Article List

Search this Journal:
Volume 14: 1 Issue (2023)
Volume 13: 4 Issues (2022): 3 Released, 1 Forthcoming
Volume 12: 4 Issues (2021)
Volume 11: 4 Issues (2020)
Volume 10: 4 Issues (2019)
Volume 9: 4 Issues (2018)
Volume 8: 4 Issues (2017)
Volume 7: 4 Issues (2016)
Volume 6: 4 Issues (2015)
Volume 5: 4 Issues (2014)
Volume 4: 4 Issues (2013)
Volume 3: 4 Issues (2012)
Volume 2: 4 Issues (2011)
Volume 1: 4 Issues (2010)
View Complete Journal Contents Listing