Knowledge based sales forecasting model for non-linear trend products

Abstract

This paper proposes a knowledge based sales forecasting approach for nonlinear trend products. Though forecasting of the future demand is an essential part of business planning and operation, most of major existing forecasting methods are applied to only regular consuming items which show linear sales trend because of seasonal cycles. In this study, using correlation between two sales-date points among similar products group, a new forecasting model is presented. It enables its accuracy advancement by incorporating experts' knowledge as grouping rules. The model is applied to Japanese literature 70 titles to verify its performance. It proved its accuracy by giving 34% error rate which is superior to 273% error rate of the exising exponential smoothing method. © Springer-Verlag London Limited 2009.