A modeling approach to multivariate analysis and clusterization theory

The paper deals with the problem of identifying the internal dependences and similarities among a large number of random processes. Linear models are considered to describe the relations among the time series, and the energy associated with the corresponding modeling error is the criterion adopted to quantify their similarities. Such an approach is interpreted in terms of graph theory suggesting a natural way to group processes together when one provides the best model to explain the others. Moreover, the clustering technique introduced in this paper will turn out to be the dynamical generalization of other multivariate procedures described in the literature.