In this work we present a novel computational approach for the extraction of underlying polymer structural component signatures and corresponding structural evolution through decomposition of multivariate X-ray scattering (SAXS/WAXS) datasets. Without assumptions based on structural geometry, this mixed-integer network component analysis (NCA) methodology generates a reduced set of component scattering signatures and component fraction evolution. Structural models are then assigned to each component based on a generalized expression for scattering from multi-phase materials. The methodology is applied systematically to the study of ethylene/alpha-olefin copolymer isothermal crystallization. The decomposition generates component signatures defining structures of varying extent within the sample but with constant average local structure. For WAXS datasets, these components can be correlated to crystalline and amorphous regions, while for SAXS datasets they can be correlated to ordered and disordered crystalline lamellae. These model choices agree with structures observed in the literature and are confirmed by comparison to reference crystallinity data. (C) 2011 Elsevier Ltd. All rights reserved.