Fuzzy matching and evidential reasoning

In this paper a general framework consisting of fuzzy database matching and evidential reasoning is presented. Data is matched onto a database in a fuzzy, i.e. quantified, way. Pieces of evidence are herefrom constructed. These update belief measures connected to the elements of the database, using a simple support belief function. A sorting and grouping of the database elements, and thresholding the beliefs, makes the process stepwise. A qualitative, unambiguous decision support is obtained at every step. The threshold and the maximum belief for a piece of evidence are the parameters varied. Some properties of the framework are examplified in a case study of identifying air targets. For a given ratio of the two parameters, the identification performance shows a surprising non-monotonicity with respect to the threshold.