Abstract
One of the foremost challenges faced by the construction industry is the issue of cost overruns. Cost overruns cut across construction projects of nations and continents as well. They vary in magnitude and occur irrespective of project size and location. Over the years numerous attempts have been made in the area of estimating cost of construction projects right and improving the efficacy or accuracy of cost estimating using different statistical methods. This research investigated the factors that contribute to cost overruns and developed a predicting cost-estimating model for public sector building projects. The aim primarily was to extract factors from historical data of completed projects and use these predictive factors to develop a predictive model. Two models were developed using the predictive variables from historical data by the use of multiple linear regression and extreme learning machine. These models were compared to see the accuracy of performance. Results from the study reveal findings that; predictive variables from historical data can be used to predict the cost of completion of construction projects at the contract award stage, the multiple linear regression model results as compared to extreme learning machine results shows that extreme learning machine performs better. The study brought to light the use of extreme learning machine for developing predicting cost-estimating models built on historical data from completed projects. This rarely exists in construction industry. It further substantiates the superior performance of extreme learning machine to multiple linear regressions using big data. The developed model can also be converted to desktop software for predicting completion cost by industry...
Ph.D. (Engineering Management)