Optimierungsbasierte Verfahren für die kurzfristige Bedarfsvorhersage in der Hochtechnologiebranche

In this paper, we study a forecast method for demand quantities in the high-technology industry. We formulate and test a parameter-driven scheme to generate forecasts. We forecast demand quantities for several future periods using the formula F = aX + (1a)Y where a is a real number from [0,1] to be estimated, X is a simple exponential smoothing forecast based on historical data, and Y are the firm orders that have been placed for the product for the current and some future months. The parameter a, and the exponential smoothing parameters are chosen to optimize a statistical measure of forecast quality. We develop a spreadsheet model to address this problem. We suggest several forecast accuracy performance measures. We present the results of computational experiments on real-world data that clearly demonstrate that the suggested approach outperforms the current forecast approach.