Ranking Model Applying Self-Organizing Maps and Factor Analysis

The present work aims to use Self-Organizing Maps (SOM) to group ranking results obtained by factor analysis. In this paper we propose the application of Factor Analysis at a stage prior to the application of the SOM, in order to use the factor scores of latent variables as input data for network training, as these scores carry practically all information regarding the variables of the lattice problem, losing only an insignificant portion of variability. Thus, the resulting map will present the cluster according to the efficiency shown in the Factor Analysis. The application refers to the performance evaluation of 50 employees of a company by means of scores on four psychological assessment tests (exams) and three indicators of sales results. The result of ranking these vendors according to performance was grouped into four homogeneous clusters.