ADVANCED DATA PREPROCESSING USING FUZZY CLUSTERING-TECHNIQUES

In this paper we describe a method of advanced data preprocessing for pattern classification using a fuzzy classification system (FCS). Inpu t pattern vectors are transformed using a matrix before classification to improve classification results and to simplify the FCS. The transf ormation matrix is calculated using the prototypes obtained with fuzzy clustering methods. The data preprocessed in this way are often class ifiable more easily than the original data. Results obtained with the new method using artificially created data and the well-known Iris dat a set are presented.