Service parts demand forecasting method for complex equipment

At present, the methods of service parts demand forecasting have many defects, such as improper input data and ignoring the relationship between historical data and the equipments' characteristics. Considering the characteristics of industrial equipments which are large-scale air compressors in this paper, historical demand data of forecasting objective and related service parts with high correlation degree were pretreated (eliminating random items and supplementing characteristics data). And a forecasting model of service parts demand was presented with BP neural network. Through the forecasting results comparison between raw data and pretreated data, it could be found that the phenomena of under-fitting vanished and the average deviation rate remarkably reduced. The method proposed in this paper has important extended significances and good portability. ICIC International © 2011 ISSN 1881-803X.