Forecasting Cost of Civil Engineering Projects Using ANN and EVA

Compared to the preliminary forecasted values, the fluctuation in the final cost and time of civil engineering projects is an issue exists in all of the countries in the world. These fluctuations will possibly result in many claims; therefore, they should be controlled. Forecasting is an important tool for controlling the fluctuations in the cost and time of a project. One of the standard methods for forecasting the time and cost of a project is the earned value analysis; also, modeling with the help of artificial neural networks is another standard forecasting method which is currently popular. In this paper, a model is created by considering several variables from the basic variables and the performance indices of the EVA as input variables without using any type of coding and only using the nftool in Matrix Laboratory software. Without any affiliation on the type of project, this model is capable of forecasting the cost of projects whose cost is higher than the budget (most of the civil engineering projects executed in Iran); the model does so with a higher accuracy compared to the cost forecast relations in the EVA. It should be noted that two water transmission channel projects and two dam projects (executed in Iran) were used to experiment this model. The findings of this experiment showed that the model is highly accurate.