A Pattern Recognition Model for Forecasting

Corporate planning has been increasingly assisted by advances in the three major areas of economic forecasting: (1) econometric models, (2) time-series analysis, and (3) business cycle indicators. The purpose of this article is to examine the effect of pre-classifying economic observations into more homogeneous groups ("patterns"). It will be shown that the formation of these groups adds further power to our more standard techniques of statistical inference. Early work in pattern recognition began with the development of computer programs for playing chess and checkers. After a brief description of pattern-recognition methodology, a procedure will be presented for pre-classifying economic events into subgroups. Empirical testing and results will be presented, as well as the refinements of this tool for better economic forecasting and planning.