A Kemeny Distance-Based Robust Fuzzy Clustering for Preference Data

Abstract We propose two robust fuzzy clustering techniques in the context of preference rankings to group judges into homogeneous clusters even case contamination due outliers or, more generally, noisy data. The C-Medoids methods, based on same suitable exponential transformation Kemeny distance, belong different approaches and differ way they introduce fuzziness membership matrix, one “m” exponent other Shannon entropy. As far as distance is concerned, it equivalent Kendall complete but differs from latter handling tied rankings. Simulations prove that our methods are able recover natural structure groups neutralizing effect possible noises outliers. Two applications real datasets also provided.