Multi-interpretation operators and approximate classification

In this paper non-classical logical techniques are introduced to formalize the analysis of multiinterpretable observation information, in particular in approximate classification processes where information on attributes of an object is to be inferred on the basis of observable properties of the object. One frequently occurring reason for imperfect classification is when the available observations are insufficient to determine unique values for each of the attributes: a range of values may still be possible. Another often occurring reason for imperfect classification occurs when the observation information is contradictory: for some of the attributes not any value is possible. The combination of both types of imperfection is non-trivial from a standard logical perspective. To address this problem multi-interpretation operators and selection operators are introduced; these techniques generalize non-monotonic reasoning formalisms such as default logic. A specific multi-interpretation operator for approximate classification is introduced and formally analysed. On the basis of this approach, in co-operation with industry a system has been designed and implemented for the analysis of ecological monitoring information.