DECISION-SUPPORT WITH NEURAL NETWORKS IN THE MANAGEMENT OF RESEARCH-AND-DEVELOPMENT - CONCEPTS AND APPLICATION TO COST ESTIMATION

Despite the small number of applications to date, neural networks are likely to be able to contribute to decision support in selected fields of R&D management. We identify the potential of neural networks in th e application domain and compare it to 'classical' applications, such as the recognition of hand-written characters. Typical neural network architectures for R&D management tend to be simple, having law complex ity, and only a small number of training samples are generally availab le. As an example, we carry out experiments for a typical R&D manageme nt application where neural networks have to estimate the final cost o f a new product under development. It turns out that neural networks b ased on the standard backpropagation learning algorithm perform reason ably well when the ratio between highest and lowest cost is small, eve n for relatively small training set sizes. Otherwise the learning algo rithm tends to undervalue low cost levels, so that deviations between estimated cost and real cost are intolerably high. Future research wil l have to investigate a modification of the error definition of the ba ckpropagation algorithm. Finally, a number of general statements are d erived from our experience, and examples are provided where neural net works are appropriate or inappropriate in the domain of R&D management . (C) 1998 Elsevier Science B.V. All rights reserved.