Advances and Applications in Statistics
Volume 65, Issue 2, Pages 151 - 164
(December 2020) http://dx.doi.org/10.17654/AS065020151 |
|
FEATURE SELECTION AND MULTICOLLINEARITY IN SUPERVISED MODELS
Mishel Qyrana and Silvia Figini
|
Abstract: This paper introduces a novel approach to perform variables selection in predictive models. Starting from the conditioning number, a new index is derived to select the most relevant variables. The contribution of this paper is two folds: in terms of methodological development, the paper introduces a new approach for future selection; in terms of computational innovation, a new algorithm is provided to implement our proposal. Empirical evidence is achieved on a simulated and real dataset. |
Keywords and phrases: feature selection, multicollinearity detection, condition number, random matrix theory.
|
|
Number of Downloads: 336 | Number of Views: 656 |
|