Abstract

Construction duration significantly influences funding, financing and resources allocation decisions that take place early in project design development. This study attempts to develop through regression analysis highway construction duration models by incorporating relevant predictor variables having statistically significant relationship with highway completion time. Historical data of highway projects initiated and completed between 2007 and 2012 were considered to enable the collection of homogenous data in terms of time, cost and other economic variables. Three multiple regression models were developed in the form of linear, semi-log and log-log transformations. The results of the analysis showed that all the three models are statistically significant and have good fit to the data with R2 values of 0.546, 0.631 and 0.940 respectively. The performances of the models were established by measuring their prediction accuracy and goodness of fit over a test sample of 15 successful projects. The result revealed that the log-log model outperformed the other models with an average % Error of -3.64%, Maximum error of 16.2% and Mean Absolute Percent Error (MAPE) of 6.87%. These results compare favourably with past studies which have shown that traditional methods of duration estimation at early project stages have values of MAPE typically in the order 10-20%.