Modeling of reliability with possibility theory

Possibility theory aims at representing and handling uncertain information. An important property of this theory is the ability to merge different data sources in order to increase the quality of the information. Different fusion rules have been defined in the literature, each with its own advantages and drawbacks. However, these rules usually cannot deal rigorously with contradictory data. Therefore, this paper proposes a new approach, based on a fusion rule using a vector expressing the reliability of the data sources. Comparisons are made with classical fusion rules. An algorithm assessing the indexes of reliability of the data used by the new role is given, and an index of the quality of the result is proposed. These three tools define a new method dealing with the reliability of the data in the fusion field, and enable a clear distinction between the data and their quality.