The paper includes some notions from the area of decision systems analysis defined for systems with multifunctions as attributes. Apart from retuned notions of indiscernibiliy relation, reduct or decision rule which are natural generalization of respective classical notions there is described an algorithm of minimal decision rules generation in considered type of decision systems. Moreover we shortly compare the rules with the ones generated as for classical decision systems. An adapted confusion matrix is presented to show output of classification of new objects to respective decision classes. We also suggest as an example a kind of real life data that are suitable for being analyzed according to the presented algorithms.