Clustering is a classification process in data mining, very used mainly for grouping of continuous values. The traditional techniques of clustering such as fuzzy C-means clustering (FCM), create groups that don't have, many times, practical sense to the user. Relative information gain has been used with success in classification applications, for instance the induction of decision tree. Our goal is to modify the way how the distance is calculated among elements in the FCM algorithm, adding to the calculation the relative information gain. The elements are grouped according to a categorical field selected from the own training dataset. Therefore groups are created and induced according to the gain criterion calculated among the elements and the categorical field.