Аннотация:A known method for finding frequent sets of attributes in the search for association rules in a binary database is based on constructing an FP-tree (Frequent Pattern Growth Tree). In the case of nonbinary data, the values of each nonbinary attribute are usually binarized using a numerical threshold, and the problem is reduced to constructing the classical FP-tree. The result significantly depends on the choice of thresholds. In this paper, it isproposed to modify the structure of the classical FP-tree by introducing an additional vertex for each nonbinary attribute;this vertex is called full, and it contains information about possible variants of binarization of this attribute values. The proposed model of the FP-tree is called full FP-tree. The application of the full FP-tree for the analysis of nonbinary data under the condition that partial orders on the attribute values are specified. Illustrative and generated examples are discussed.