Fuzzy partition-based distance practical use and implementation

Abstract

This work discusses the implementation of a semidistance based on fuzzy partitions, that allows to introduce expert knowledge into distance computations done on numerical data. It can be used in various kinds of statistical clustering or other applications. The semi-distance univariate behaviour is first studied, then a multivariate clustering case study is presented, that includes an optimization procedure. All calculations are done using open source softwares: FisPro for fuzzy modelling and R for clustering. Emphasis is given to a number of software functionalities, which are targeted at the practical use of the semi-distance.