Deep Learning Generalized Structured Component Analysis: An Interpretable Artificial Neural Network Model with Composite Indexes

DL-GSCA integrates neural networks with generalized structured component analysis so that composite indexes can capture nonlinear relationships among their indicators. The approach learns these nonlinear functions without requiring researchers to specify their form in advance, while preserving an interpretable network among the components. Simulation and empirical examples assess whether this flexibility improves predictive performance relative to conventional GSCA when indicator relationships are nonlinear.