TY - JOUR
T1 - A framework for big data analytics approach to failure prediction of construction firms
AU - Alaka, Hafiz
AU - Oyedele, Lukumon O.
AU - Owolabi, Hakeem A
AU - Bilal, Muhammad
AU - Ajayi, Saheed O.
AU - Akinade, Olugbenga O.
N1 - © 2018 The Author(s). Production and hosting by Elsevier B.V. on behalf of King Saud University. This manuscript is made available under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International licence (CC BY-NC-ND 4.0). For further details please see: https://creativecommons.org/licenses/by-nc-nd/4.0/
PY - 2018/4/12
Y1 - 2018/4/12
N2 - This study explored use of big data analytics (BDA) to analyse data of a large number of construction firms to develop a construction business failure prediction model (CB-FPM). Careful analysis of literature revealed financial ratios as the best form of variable for this problem. Because of MapReduce’s unsuitability for iteration problems involved in developing CB-FPMs, various BDA initiatives for iteration problems were identified. A BDA framework for developing CB-FPM was proposed. It was validated by using 150,000 datacells of 30,000 construction firms, artificial neural network, Amazon Elastic Compute Cloud, Apache Spark and the R software. The BDA CB-FPM was developed in eight seconds while the same process without BDA was aborted after nine hours without success. This shows the issue of not wanting to use large dataset to develop CB-FPM due to tedious duration is resolvable by applying BDA technique. The BDA CB-FPM largely outperformed an ordinary CB-FPM developed with a dataset of 200 construction firms, proving that use of larger sample size with the aid of BDA, leads to better performing CB-FPMs. The high financial and social cost associated with misclassifications (i.e. model error) thus makes adoption of BDA CB-FPMs very important for, among others, financiers, clients and policy makers.
AB - This study explored use of big data analytics (BDA) to analyse data of a large number of construction firms to develop a construction business failure prediction model (CB-FPM). Careful analysis of literature revealed financial ratios as the best form of variable for this problem. Because of MapReduce’s unsuitability for iteration problems involved in developing CB-FPMs, various BDA initiatives for iteration problems were identified. A BDA framework for developing CB-FPM was proposed. It was validated by using 150,000 datacells of 30,000 construction firms, artificial neural network, Amazon Elastic Compute Cloud, Apache Spark and the R software. The BDA CB-FPM was developed in eight seconds while the same process without BDA was aborted after nine hours without success. This shows the issue of not wanting to use large dataset to develop CB-FPM due to tedious duration is resolvable by applying BDA technique. The BDA CB-FPM largely outperformed an ordinary CB-FPM developed with a dataset of 200 construction firms, proving that use of larger sample size with the aid of BDA, leads to better performing CB-FPMs. The high financial and social cost associated with misclassifications (i.e. model error) thus makes adoption of BDA CB-FPMs very important for, among others, financiers, clients and policy makers.
KW - Big data analytics
KW - failure prediction models
KW - construction businesses
KW - machine learning
KW - MapReduce/Spark
U2 - 10.1016/j.aci.2018.04.003
DO - 10.1016/j.aci.2018.04.003
M3 - Article
SN - 2634-1964
JO - Applied Computing and Informatics (ACI)
JF - Applied Computing and Informatics (ACI)
ER -