Using nonstationary fuzzy sets to improve the tractability of fuzzy association rules

Modern organisations now collect very large volumes of data about customers, suppliers and other factors which may impact upon their business. There is a clear need to be able to mine this data and present it to decision makers in a clear and coherent manner. Fuzzy association rules are a popular method to identifying important and meaningful relationships within large data sets. Recently a fuzzy association rule has been proposed that uses the 2-tuple linguistic representation. This paper presents a methodology which makes use of non-stationary fuzzy sets to post process 2-tuple fuzzy association rules reducing the size of the mined rule set by around 20% whilst retaining the semantic meaning of the rule set.