A neural network for signal modelling in business cycle studies

Business cycle study demands simultaneous investigation of several economic indicators. The state space formulation is adopted here for analyzing multiple time series data to this end, and a neural network is employed to characterize their common growth dynamics. For demonstration, four key variables for the USA are considered, viz., sales, production, unemployment and personal income. Their monthly data over 1965-1989 are examined. The analysis encompasses the identification of the business cycle turning points, i.e., the dates of recessions.<
>