Functional Dimension Reduction in Predictive Modeling
Evgeny Vladimirovich Burnaev, Alexander Bernstein · Journal of Communications Technology and Electronics · 2021
Abstract The paper considers the problem of dimension reduction in metamodeling (predictive modeling). It is shown that to construct an exact metamodel based on reduced data, it is necessary to solve the problem of dimension reduction in a nonstandard formulation with due account of the additional functional constraints. To construct a tangent affine linear subspace from the data, which is required for solving the functional dimension reduction problem, a generalized principal component construction problem is formulated and solved.