OVERVIEW: Curve Fitting is actually a set of techniques particularly appropriate for very new technologies with little or no data, or very long range forecasting. These are highly exploratory methods and large errors are quite likely.
BASIC IDEA: With a few data points, subjective and freehand extrapolation of data. Curve fitting is done to approximate the basic trend so that by extrapolating the trend, a forecast can be made.
PROCEDURE: The simplest curve to fit to a few data points is a straight line. The forecaster then interacts with this type of mechanical extrapolation to incorporate technical judgment in order to make a forecast. Note that the use of a straight line implies that the overall trend would increase indefinitely. Nevertheless, a simple linear trend may be quite reasonable, especially when looking at a class of technologies.
EXAMPLE: An example of using a simple (linear) trend extrapolation is shown below for forecasting the areal densities for computer rigid disks. Note that such a simple trend cannot be extrapolated indefinitely because the curve would increase to infinity. Thus, theoretical limits of the technology would suggest that the trend will have to bend a some point.
Another type of trend is an S-Shaped Curve.