HYBRID PATTERN RECOGNITION METHOD TO DIAGNOSE DYNAMIC SYSTEMS

We use the supervised classification method Fuzzy Pattern Matching (FPM) to realize the diagnosis of dynamic systems. FPM is decentralized, i.e., its global decision is based on the selection of one of the intermediate decisions. Each intermediate decision is based on one attribute. Thus, FPM does not take into account the correlation between attributes. Additionally, FPM does not respect the shape of classes if this shape is non convex. These drawbacks can be solved by using the expert knowledge based on a set of rules which provide the information about attributes correlation and the shape of classes. However this information is imprecise. Thus in this paper, we propose to combine FPM with the expert rules in order to obtain efficient and precise decision about the class of new patterns. We call this combination Hybrid FPM (HFPM). Three examples are used to show the performances of HFPM with respect to the classical FPM.