抄録
A method of forecasting shipping number of parts is proposed for build-to-order products using quote information in pre-sales as an advance indicator. As the target shipping number of parts changes irregularity in a short cycle, it is difficult to forecast with conventional methods. In case of build-to-order product, quote information is useful for forecasting multiplying by order probability. But the problem is how to identify the order probability includes unobservable noise. The method assumes the noise as normal random number and corrects it with Kalman filter. The shipping number is calculated using identified order probability and previous quoted information. The method was evaluated using 78 types of parts of an information equipment product. Experimental result demonstrates that shipping number of parts whose average rate of change are less than 0.4 can be estimated with 28.2% of mean absolute percent error, making the method available for forecasting the target shipping number.
