Abstract
The proposed clustering algorithm is aimed at revealing the structure within the patterns under a simultaneous satisfaction of directionality constraints. These constraints are utilized to cope with functional relationships between the specified features of the patterns. To address this aspect of directionality, the introduced objective function (clustering criterion) is asymmetric. Proposed is also a criterion determining a “plausible” number of clusters within the data set. The instantaneous use of the outcomes of clustering for system identification will be also revealed. Numerical studies use synthetic and experimental data.