A NEURAL NETWORK APPROACH TO FORECASTING INVENTORIES FROM INTEREST RATES

摘要

Inventory investment, which is 0.6% of gross national product (GNP), accounts for about half the size of the changes in GNP and inverted interest rates are predictors of economy activity. Business people see evidence that interest rates affect inventory behavior. Nevertheless, the interest rates inventory relationship remains an important open and difficult research problem. Therefore, we propose to use a time-delayed neural network trained with the Levenberg-Marquardt backpropagation algorithm to determine the relationship between inverted nominal interest rates and aggregate inventory as well as to then predict inventory behavior from interest rates. We did not achieve a unique prediction. We achieved many statistically accurate predictions at the 95% confidence level. Our inconclusive result suggests a complex interest rates aggregate inventory relationship that might be resolved by studying the relationship between interest rates and inventories' components.