This thesis considers expert knowledge modelling by implicative fuzzy rules. It explores the benefits of these rules compared to the most frequently used fuzzy rules: conjunctive rules.
However, inference from implicative rules and fuzzy inputs is not easy and has long been an impediment to their use. The main difficulties are the complexity of the inference with several implicative rules and fuzzy inputs, the partition design, and the semantic interpretation for users familiar with the reasoning with conjunctive fuzzy rules.
Our work focuses on these points. We present an inference method using implicative fuzzy rules and fuzzy inputs, which can easily implement the implicative reasoning in the one and two-dimensional case.
We also give a comparison between conjunctive rules and implicative rules, and we study the semantics of these rules, in terms of logic and practical use.
A real world illustration in the food industry is presented. The goal of this work is the prediction of post maturing cheese defects with information available before the maturing process. Available information comes from CTFC (Technical Center on Comtois Cheese) expert knowledge and process data.
Since the developed methods are generic, they can be used for a wide class of applications: those in which the expert knowledge is expressed in the form of a model. They provide modeling perspectives that respect both imprecise data and expert reasoning characteristics.