It is often useful to incorporate a metamodel into the sensitivity analysis of the economics of capital projects. While statistical models are straightforward and understandable, they are not always appropriate, especially where there are discontinuities or complex relationships in the economic data. This paper puts forth a neural network approach that uses design of experiments to generate metamodel training and validation data from the project(s) under consideration. Both backpropagation multi-layered perceptron networks and cascade correlation networks are explored and compared with the statistical approach. It is shown that the backpropagation networks work well for a variety of data sets and can be superior to regression in prediction accuracy when the economic data is especially complicated.Evaluation and selection of capital investments are generally based on an economic measure such as net present value (NPV), which is calculated from the estimates of input factors such as interest rate and amount and timing of cash flows. An investment decision made on the basis of point estimates of these input factors may not be robust. Sensitivity analysis thus becomes crucial as it allows the decision maker to determine the relative effect of changes in the input estimates on the indicated decision. Sensitivity analysis will also identify the factors that are most influential to the outcome of the investment decision.Precise estimates of important input factors are critical to a correct decision. Moreover, it will be beneficial to monitor these factors during implementation of Downloaded by [North West University] at 12:13 22 December 2014 the investment so that the decision maker can take corrective action in a timely manner.One factor at a time and scenario generation approaches are popular sensitivity analysis techniques. The one factor at a time approach is carried out by assuming a likely range or a single value for each input factor and recalculating tbe economic measure. The results are then presented in a graphical or tabular format (see for example [S]). Drawbacks to this approach are that it does not consider interactions among factors and it becomes more difficult to portray as the number of factors increases. Scenario generation explores the effect on the economic criterion when different factors take on the most likely, optimistic and pessimistic estimates in various combinations [3].
,The results are usually tabulated. However, the size of the matrix display in the table grows quickly, thereby complicating the comprehension of useful information when multiple factors are analyzed.A metamodel is a simplified or approximated descriptive model of another descriptive model. It, itself, is not prescriptive, however it can assist with that function using optimization algorithms or interactive user queries. It is constructed from input-output data gathered from calculation, simulation or experimentation. Design of experiments is often used to collect data so that a meaningful metamodel is obtained in an efici...