Polynomial forecasting utilizing exponential smoothing on successive coefficient determinations

Economic forecasting techniques are being successfully applied to problems of inventory and production control. Literature on this subject is generally divided into two classifications: the first being a qualitative approach dealing with business cycles or long-term economic growth, and the second dealing with methods of obtaining short-term forecasts of product demand. Among the techniques for the latter group are the exponential smoothing models developed by R. G. Brown. These methods provide reasonably accurate and reliable forecasts. In addition, they require a minimum of computations and are suitable for use with digital computers. Two new models are developed here that use the algorithm of single exponential smoothing to weigh the coefficients of a polynomial such that, for each new data point, a new equation is obtained. These models retain all the advantages of exponential smoothing while providing a new approach for predicting future occurrences. Since there are many situations which can be adequately forecasted using linear trends, a model is presented that makes use of the equation for a straight line. The coefficients of this equation are successively smoothed in a sequential manner throughout the time series. The result is a model in which the most recent information is exponentially weighted based on a smoothing constant. ..