A supervised variational autoencoder framework for dimensionality reduction and predictive modeling in high-dimensional socioeconomic data

We introduce an estimation framework utilizing a Supervised Variational Autoencoder (SVAE) to address challenges posed by high-dimensional socioeconomic data. Unlike classical linear dimensionality reduction methods, such as PCA and Lasso regression, the proposed SVAE effectively captures complex no...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Journal of Economy and Technology Jg. 4; S. 9 - 19
Hauptverfasser: Xue, Pei, Li, Tianshun
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier B.V 2026
KeAi Communications Co., Ltd
Schlagworte:
Dimensionality reduction
High-dimensional data
Nonlinear representation learning
Predictive modeling
Socioeconomic forecasting
Supervised variational autoencoder
Dimensionality reduction
High-dimensional data
Socioeconomic forecasting
Predictive modeling
Nonlinear representation learning
Supervised variational autoencoder
ISSN:2949-9488, 2949-9488
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:We introduce an estimation framework utilizing a Supervised Variational Autoencoder (SVAE) to address challenges posed by high-dimensional socioeconomic data. Unlike classical linear dimensionality reduction methods, such as PCA and Lasso regression, the proposed SVAE effectively captures complex nonlinear interactions through supervised latent representation learning. Empirical analyses using comprehensive cross-country data from the World Bank (196 countries, 1997–2023) demonstrate the SVAE framework’s superior predictive accuracy, interpretability, and robustness in forecasting GDP growth, highlighting its potential for policy evaluation and macroeconomic forecasting.
ISSN:2949-9488
2949-9488
DOI:10.1016/j.ject.2025.06.001