Possibilistic missing data estimation.
Résumé
An approach that deals with the heterogeneity and the imperfection of information elements which constitute the objects in large databases has been proposed in this paper. Unlike the prior works that separately tackle these aspects using complex and conditional techniques, our method is general and takes account of them within a simple, flexible, and robust unified framework. It is fundamentally based on two fuzzy monotone measures: the possibility and the necessity degrees introduced in the theory of possibilities. A simple concrete example will also be given to clarify and to simply illustrate the main steps of computation, pointing out the outperformance and the robustness of the proposed strategy.