SERFIT - AN ALGORITHM TO FORECAST MINERAL TRENDS

A PASCAL computer program, named SERFIT, facilitates the identificatio n of trend model for long-term forecasting and the estimation of model parameters. Model identification is achieved through the computation of slope characteristics from mineral data time series. The trend mode ls generated by the program are: linear, normal, lognormal, and modifi ed exponentials: simple-modified exponential, logistic, derivative log istic, Gompertz, and derivative Gompertz. Parameters of the family of modified exponential models are estimated using Mitscherlich's regress ion, which is based upon the maximum likelihood method and provides a probability structure for the models. SERFIT is demonstrated on U.S. p etroleum production and world copper consumption data.