Context-based information quality for sequential decision making

Abstract

The context-dependent nature of information quality has been well recognized in the literature. Depending on the context, the overall information quality may relate to a single quality attribute, a combination of several, or all the attributes. Selection of quality attributes, methods for their combination as well as quality control strategies require consideration of decision makers' objectives and functions together with context parameters such as characteristics of the environment and current situations. This paper considers such context-dependent information quality attributes, namely, credibility, reliability, and timeliness, and presents a method of incorporating their combination into sequential decision making for pattern recognition. The context is represented by the time-dependent distance between an observed target and a sensor, and a situation-based time-dependent threshold on credibility. A case study involving an evidential learning system for sequential decision making designed for a quantitative evaluation of the described method, is also presented.