Adaptive dynamic input‐output analysis using neural networks: Japanese industrial structure

Considers an application of adaptive control policy to dynamic input‐output systems of Japanese large‐scale industrial (primary, secondary and tertiary) sectors by neural networks. The adaptive control policy has three steps. The first is to obtain the optimal control policy such that the minimization of the weighted sum of the squared deviation between the actual targets and the desired subject to econometric models is achieved. The second is to determine the optimal outputs for each industrial sector through dynamic input‐output system under the optimal control policies. The third is to obtain the network outputs by neural network algorithm through the controlled output equations derived from DIO system. We consider what affects the outputs if the optimal control policy was adopted, and how the change of industrial structure has occurred after the bubble burst in 1990s in Japan during 1985 through 1993, and we predict the future of the industries up to 2010 by using DIO linked to the final demand econometric models of the Japanese industrial sectors by simulation.