[2I5-GS-10] AI application: Industry

[2I5-GS-10-03] Demand Forecasting for Electronic Devices Considering Uncertainty of Order and Production

〇Jeffry Nainggolan¹, Takuto Ichikawa¹, Koji Miura¹, Yosuke Tajika¹ (1. Panasonic Industry Co., Ltd.)

Keywords:Demand forecasting

To predict future demand, generally past sales are used. In order to improve the accuracy of demand forecasting, we propose a forecast model that considers planned number of shipments in addition to forecast based on past sales. In this model, the planned number of shipments is corrected by weight α1 which represents its uncertainty. Similarly, the forecast based on past sales is corrected by weight α2 which represents its uncertainty. In forecasting demand for electronic devices up to 12 months in advance, the forecast error of the proposed method is reduced by 30% compared to the forecast error based on past sales.