Extracting Fuzzy Classifier Rules from Mixed Continuous and Categorical Data

In many applications, data objects are described by both continuous and categorical features. This study presents a new method for generating a fuzzy rule base from mixed continuous and categorical data. Basic fuzzy classifiers can only be applied to continuous attribute values and cannot handle mixed data. k-means and k-prototype algorithms were used for clustering. Three rule extraction methods are considered: a method based on categorical feature frequency (FM), a method based on ordinal encoding (OC), and a method based on hyperbox (HM). The proposed methods were tested on datasets from the KEEL database. The FM method showed the highest classification accuracy on most data sets.