In deterministic identification the identified system is determined on
the basis of a complexity measure of models and a misfit measure of
models with respect to data. The choice of these measures and
corresponding notions of optimality depend on the objectives of
modelling. In this monograph, the cases of exact modelling, model
reduction and approximate modelling are investigated. For the case of
exact modelling a procedure is presented which is inspired by objectives
of simplicity and corroboration. This procedure also gives a new
solution for the partial realization problem. Further, appealing
measures of complexity and distance for linear systems are defined and
explicit numerical expressions are derived. A simple and new procedure
for approximating a given system by one of less complexity is described.
Finally, procedures and algorithms for deterministic time series
analysis are presented. The procedures and algorithms are illustrated by
simple examples and by numerical simulations.