Grey forecast model for torpedo development cost based on neural network

Abstract

Cumulative curve of development cost for torpedo shows similar to shape S. Then grey Verhulst model of development cost for torpedo can be established by collecting the development cost historical data for several years. The fitted values and the predictive value can be gotten. To take full advantage of information contains in the historical data, a mapping between historical data and the predictive value using BP neural networks can be found through BP Neural Networks. Example shows that this method has good prediction accuracy.