Selecting a "best subset" of input variables is a critical issue in forecasting. This is especially true when the number of available input series is large, and an exhaustive search through all combinations of variables is computationally infeasible. Inclusion of irrelevant variables not only doesn't help prediction, but can reduce forecasting accuracy through added noise or systematic bias. We demonstrate a technique called "sensitivity-based pruning" (SBP) that removes irrelevant input variables from a nonlinear forecasting or regression model. The technique makes use of a saliency measure computed for each input variable and uses estimates of prediction risk for determining the number of input variables to prune. We present preliminary results of the SBP technique applied to neural network predictors of a key business cycle measure, the US Index of Industrial Production.

INKA is a natural language interface to facilitate knowledge acquisition during expert system development for electronic instrument trouble-shooting. The expert system design methodology develops a domain definition, called GLIB, in the form of a semantic grammar. This grammar format enables GLIB to be used with the INGLISH interface, which constrains users to create statements within a subset of English. Incremental parsing in INGLISH allows immediate remedial information to be generated if a user deviates from the sublanguage. Sentences are translated into production rules using the methodology of lexical-functional grammar. The system is written in Smalltalk and, in INKA, produces rules for a Prolog inference engine.