Modelling Cost-Risk Impact on Nigerian Highway Projects Using Multiple Linear Regression and Artificial Neural Networks

Abstract

Cost predictive models for highway projects are relatively scarce in developing countries, despite the frequency and magnitude of project cost overruns in such countries. This study identified critical cost risks impacting highway projects and modelled their impacts on the actual cost of the projects. Historical cost data on highway projects published by the Nigerian Federal Ministry of Power, Works and Housing in 2017 served as a preliminary list of projects for the study, while further cost data were obtained from highway engineers and quantity surveyors across Nigeria using the snowballing technique until 103 highway projects were identified. Project participants were purposively chosen to fill out questionnaires on costrisks factors associated with highway construction projects. The relative importance index and Pareto 80/20 rule were used to analyse the collected primary data. Thereafter, multiple linear regression and artificial neural network models were developed. Findings revealed that the increase in the cost of construction materials and labour and the fluctuations in foreign exchange rates were the most significant risks impacting highway project cost performance. A comparison of the models indicated that the artificial neural networks model performed better. Hence, the artificial neural networks model is a superior technique for modelling the relationship between cost risks and cost performance of highway projects.