Le Manager
Volume 12, Numéro 2, Pages 595-631
2025-12-25
Cross Validation Across Regularized Linear Regression
Authors : Mohamed Bachir Aoun . Fatiha Mezoued .
Abstract
This study evaluates various regression model selection methods with multi-dimensional settings, using parameters from (9)for low dimensional parameters and (12) for high dimensional setup. The experiments varies sample sizes, variables number, correlation levels, and Gaussian noise. The assessment covers LASSO, Ridge, Elastic-Net, MCP and SCAD as sparse regularization methods, adaptive methods and stepwise selection methods. Performance is evaluated based on the cross validation criterion as GCV, V-Fold Cross Validation CV-V and LOOCV. In low dimensional data, LOOCV outperform the other criterion. For high dimensional generation, CV5, CV10, and GCV are preferred depending on the model and correlation strength. The findings offer clarity for model selection and encourage future research into real-world applications across diverse fields.
Keywords
High dimension ; Linear Regression ; Sparse regularization methods ; Cross Validation ; Model selection ; Evaluation Criteria
Les articles similaires
Mohamed Bachir Aoun
.
Fatiha Mezoued
.
pages 31-46.
Abdelhakim Necir
.
Brahimi Brahim
.
pages 03-15.
Zemri Bouazza Elamine
.
Khetib Sidi Mohamed
.
pages 624-637.
Boutaghane Mohammed Abderrezak
.
Yekhlef Abdellah
.
pages 194-214.
Medini Atmane
.
Aouadi :mostafa
.
pages 437-448.
