Réflexions autour d’un indice de confiance pour l’aide à la décision sous information incomplète

In some application contexts, decision-making can be made difficult by the absence of certain information. This information may be difficult to produce (e.g., expensive), or difficult to gather (e.g., requiring a survey). In this article, we describe our thoughts on developing a fuzzy rule base adapted to this context. Some rules will need to be able to activate despite the absence of a significant number of inputs to produce a result. Since some rules will not be activated, we propose to accompany the inference result with a confidence index. This reflects both the importance of the data present and the convergence of the activated rules to the produced output.