A simplified linear predictor with guaranteed error

This paper presents a technique for obtaining a predictor of the output of a complex linear system. The predicted output is obtained by choosing a suitable combination of some variables which can represent the "most important" states of the system, in such a way that an easily computable bound of the prediction error is minimized. Such a bound moreover results very useful since it gives a measure of the goodness in the choice of the "important" states. The proposed approach seems to be attractive for its simplicity and for the fact that a bound on the prediction error can be estimated.