Interpretable Approaches for Fuzzy Interpolative Reasoning


Fuzzy reasoning methods are often based on fuzzy rules, easily understandable by users. In this paper, we analyse the case of an incomplete rule base, requiring an interpolation between successive rules. We first present a list of interpretability criteria of fuzzy interpolative methods studied in the literature, in relation with Explainable Artificial Intelligence. We study in particular methods based on analogy and we propose a new method of fuzzy interpolative reasoning whose model and conclusion are easily interpretable. An experimental comparison with existing methods is then proposed to highlight the qualities of interpretability of this new method.