Proceedings of International Structural Engineering and Construction
Holistic Overview of Structural Design and Construction
Edited by Vacanas, Y., Danezis, C., M., Yazdani, S., and Singh, A.
Copyright © 2020 ISEC Press
ISSN: 2644-108X, ISBN: 978-0-9960437-8-6
CON-16-1
FORECASTING CONSTRUCTION DELAY TIMES IN HIGH-
RISE BUILDING PROJECTS
MUIZZ SANNI-ANIBIRE
1
, MOHAMAD ZIN
2
and SUNDAY OLATUNJI
3
1
Dammam Community College,
King Fahd University of Pet. And Min., Dammam, Saudi Arabia
2
School of Civil Engineering, Faculty of Engineering,
Universiti Teknologi Malaysia, Johor Bahru, Malaysia
3
Department of Computer Science, College of Computer Science and Information Technology,
Imam Abdulrahman Bin Faisal University, Saudi Arabia
High-rise buildings, which have become a significant part of the urban habitat, is particularly notorious for
their delayed completion times. Though, there exists a plethora of studies on construction delays, the problem
however is insufficient research on prescriptive methods to mitigate delays. Thus, this study sought to employ
Machine Learning (ML) techniques to learn from historical data on high-rise construction to forecast
potential delay times. An input data containing 9 features, and 12 cases was used. Initially five feature sets
were built based on the recursive feature elimination process. Further to that was the classification process
that employs the following ML techniques: Multi-Linear Regression Analysis (MLRA), k-Nearest Neighbors
(KNN), Artificial Neural Networks (ANN), and Support Vector Machines (SVM) to determine delay times.
The predictive performance of these techniques was measured using their correlation co-efficient (R
2
) and
their Root Mean Squared Errors (RMSE). The best three models according to the ML techniques used was
SVM with 2 features (R
2
0.56, RMSE 1.6), ANN with 2 features (R
2
0.49, RMSE 1.83), and KNN with all
features (R
2
0.46, RMSE 1.71). To seek improvement of the predictive performance of the models developed,
the three best performing models were combined using fixed and trained rules. The results showed an
improvement for a fixed rule based on the minimum values with (R
2
0.59, RMSE 1.65). The study has
significant implications in the risk management process of high-rise projects to avoid delays. The originality
is evident in that this is the first study that employs ML in predicting construction delay times.
Keywords: Delay, Artificial Intelligence, Machine Learning, High-rise.
1 INTRODUCTION
The 21st century is witnessing a rising complexity in buildings, observable in the rapid growth of tall
buildings in urban centers globally. Despite the potential of tall buildings in the urban context to be a
sustainable solution to an impeding housing crisis, such mega structures are subject to underperformance
issues that already plague the industry. Significant among these is the problem of delays, which
consequently leads to time overruns, cost overruns, dispute, arbitration and litigation, total abandonment
and dissatisfied stakeholders. Interestingly, CTBUH (2014) in its report “Dream Deferred: Unfinished Tall
Buildings” noted the alarming rate of increase of “never completed” tall buildings. Though, there exists a
plethora of studies on the subject of construction delay in the research arena, the problem is that these
studies tend to be descriptive and exploratory, and thus inadequate in providing the desired solution sought
by the industry. Remarkably, the construction industry in its bid to solve its unproductivity issues is turning
towards computers and automation. Recent efforts towards automation can be observed in some of the
studies that seek to adopt Artificial Intelligence (Attal, 2010; Czarnigowska and Sobotka, 2014; Bayram,
2017; Peško et al., 2017). These studies have been directed towards the estimation of costs and duration of
construction projects. No study, to the best of the authors’ knowledge, has sought to address the
estimation/prediction of construction delay times, which may be attributed to the unavailability of relevant