Abstraction mechanisms in discrete-event inductive modeling

The power of abstraction lies in its ability to deal with 'lack' of knowledge. In this regard, success in modeling and simulation rests on discovering useful abstractions that can support objectives of modeling. In our treatment, we refer to 'data abstraction' as opposed to 'structure simplification' since we consider a system's behavior rather than its structure. A system's behavior can be represented as time-varying input/output segments. Given the behavior of a causal, time-invariant system, we define some basic abstraction mechanisms to support inductive modeling. The basis for these abstraction mechanisms are a set of general assumptions which allow consistent abstraction of IO segments. Then, given these assumptions and non-monotonic reasoning paradigm, capable of handling them, we try to tackle the fundamental problem of insufficient knowledge in the realm of inductive modeling. In this way, by making useful abstractions, we can predict a system's unobserved behavior according to a well-defined framework of discrete-event inductive modeling.