This paper deals with the creation of multidimensional data classes – macrosituations – by decreasing their dimension. A large number of monitored attributes in examined situations in complex systems often complicates technical realization of classification and extends the time needed for providing a decision. It is possible to decrease the dimension of situations and, simultaneously, to not decrease decision-making quality. The main subject relates to a possible approach – the Principal Component Method. The basis of this method lies in finding a linear transformation of original p-dimensional space of attributes into a new p’-dimensional space of attributes where p’≤p. New attributes, called principal components, arise in a suitable linear combination of original attributes and are sorted in descending order based on their variance.