This two-phase analysis method combines the ease of use of self-organizing maps and robustness of statistical methods. Groups of data records having similar properties or indicating similar behavior can be located easily and identified using prior knowledge. Conclusions are obtained in linguistic form, and clustering error estimates are included in the results.
Main advantages to statistical methods, for example factor analysis combined with the use of an arithmetic mean of each category, are the ability to locate special behavioral groups within each category and clearness of the results. Results can be verified throughly by calculating the number of misclassifications.