PREDICTIVE MODELLING OF ROAD CONSTRUCTION COST: CASE STUDY OF ONDO STATE, NIGERIA

Abstract

Non-linear regression models have been calibrated to predict the cost of road construction based on data collected from contracts awarded for some of the past road construction projects in Ondo State. Models were based on data collected for70 projects undertakenbetween 2005-2012.The modelcomponents include those for Earthworks, Sub-base, base coarse and asphalt works. A sub-model for each activity was developed based on the used dependent variables which are total cost of project, the cost of earthworks, the cost of base work, the cost of asphalt work and the contigencies. Independent variables are road quantities: road length, road width and the thickness.Also the base works, earthwork (cut and fill) and the asphalt works were considered as independent variables. The coefficient of determination (R2) of the calibratedmodels ranged from 0.79 to 0.96. TheR2for the total project cost and asphalt gives 0.955 and it impliesthat the predictor (quantity of asphalt work) accounted for 95.5% of the dependent variable (total project cost) in the study area, while the remaining 4.5% was due to other factors that are not captured in this model as accounted for by the error term (Ԑ). The adjusted R-squared of 0.952 implied that 95.2% of the variations in the project cost are explained by only those independent variables that have significant effect on the project cost (dependent variable) in the study area.It was alsorevealedthat the coefficient of asphalt work quantity had a significant effect on the total project cost at 5% level of significance with the t-value of the road width variable not up to 2 which represents the level of its non-significance in the model. The seven models calibrated were finally tested using Mean Absolute Percentage Error (MAPE) and they were discovered to be helpful insavingprofessional’s time, making more realistic decisions and also preventing the underestimating and overestimating of project costs.