APPLICATION OF THE FUZZY-C-MEANS ALGORITHM TO THE STUDY OF RELATIONSHIPS BETWEEN COMPOSITION AND PROPERTIES - PHYSICAL-PROPERTIES OF GLASSES

The opportunities of applying the Fuzzy-C-Means (FCM) algorithm for cl uster analysis to study composition-property relations are discussed. Physical properties of 98 aluminosilicate glasses serve as an example. The influences of the initial partition and of the number of clusters chosen are discussed in detail. The FCM algorithm allows to find char acteristic members of clusters (these form the ''nuclei'' of the clust ers) and to find outliers of the data set under consideration. Thus th e results of the cluster analysis form a basis for a deeper understand ing of relations between the composition and properties of the glasses studied.