Visualizing Profiles of Large Datasets of Weighted and Mixed Data

This work provides a procedure with which to construct and visualize profiles, i.e., groups of individuals similar characteristics, for weighted mixed data by combining two classical multivariate techniques, multidimensional scaling (MDS) the k-prototypes clustering algorithm. The well-known drawback MDS in large datasets is circumvented selecting small random sample dataset, whose are clustered means an adapted version algorithm mapped via MDS. Gower’s interpolation formula used project remaining onto previous configuration. In all process, distance measure proximity between individuals. methodology illustrated on real obtained from Survey Health, Ageing Retirement Europe (SHARE), was carried out 19 countries represents over 124 million aged Europe. performance method evaluated through simulation study, results point that new proposal solves high computational cost low error.