Finding fuzzy descriptions using SOMs

Autoren Mario Drobics
Titel Finding fuzzy descriptions using SOMs
Typ Technischer Bericht
Nummer SCCH-TR-0008
Ort Hagenberg, Austria
Institution SCCH
Abteilung KVS
Jahr 2000

Self Organizing Maps (SOMs) are often used to analyze and visualize great datasets. Though they ca show neighborhood relation (and clusters) in a very good way, they lack a natural description of the data points. On the other hand, traditional inductive learning methods that are able to gain more descriptive results, misconduct when applied to big and noisy data.In this paper we will present an combined approach, where SOMs are used to smooth the input data and to reduce the number of samples, while high sophisticated inductive learning methods are used to create a set of fuzzy rules, to describe the dataset in a human-like manner.