LASSO and Elastic Net Tend to Over-Select Features
Machine learning methods have been a standard approach to select features that are associated with an outcome and to build a prediction model when the number of candidate features is large. LASSO is one of the most popular approaches to this end. The LASSO approach selects features with large regres...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2023-08-01
|
Series: | Mathematics |
Subjects: | |
Online Access: | https://www.mdpi.com/2227-7390/11/17/3738 |
_version_ | 1797582226375835648 |
---|---|
author | Lu Liu Junheng Gao Georgia Beasley Sin-Ho Jung |
author_facet | Lu Liu Junheng Gao Georgia Beasley Sin-Ho Jung |
author_sort | Lu Liu |
collection | DOAJ |
description | Machine learning methods have been a standard approach to select features that are associated with an outcome and to build a prediction model when the number of candidate features is large. LASSO is one of the most popular approaches to this end. The LASSO approach selects features with large regression estimates, rather than based on statistical significance, that are associated with the outcome by imposing an <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msub><mi>L</mi><mn>1</mn></msub></semantics></math></inline-formula>-norm penalty to overcome the high dimensionality of the candidate features. As a result, LASSO may select insignificant features while possibly missing significant ones. Furthermore, from our experience, LASSO has been found to select too many features. By selecting features that are not associated with the outcome, we may have to spend more cost to collect and manage them in the future use of a fitted prediction model. Using the combination of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msub><mi>L</mi><mn>1</mn></msub></semantics></math></inline-formula>- and <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msub><mi>L</mi><mn>2</mn></msub></semantics></math></inline-formula>-norm penalties, elastic net (EN) tends to select even more features than LASSO. The overly selected features that are not associated with the outcome act like white noise, so that the fitted prediction model may lose prediction accuracy. In this paper, we propose to use standard regression methods, without any penalizing approach, combined with a stepwise variable selection procedure to overcome these issues. Unlike LASSO and EN, this method selects features based on statistical significance. Through extensive simulations, we show that this maximum likelihood estimation-based method selects a very small number of features while maintaining a high prediction power, whereas LASSO and EN make a large number of false selections to result in loss of prediction accuracy. Contrary to LASSO and EN, the regression methods combined with a stepwise variable selection method is a standard statistical method, so that any biostatistician can use it to analyze high-dimensional data, even without advanced bioinformatics knowledge. |
first_indexed | 2024-03-10T23:18:00Z |
format | Article |
id | doaj.art-f6df4ebb2c3348b29bde0eccf43f9141 |
institution | Directory Open Access Journal |
issn | 2227-7390 |
language | English |
last_indexed | 2024-03-10T23:18:00Z |
publishDate | 2023-08-01 |
publisher | MDPI AG |
record_format | Article |
series | Mathematics |
spelling | doaj.art-f6df4ebb2c3348b29bde0eccf43f91412023-11-19T08:31:28ZengMDPI AGMathematics2227-73902023-08-011117373810.3390/math11173738LASSO and Elastic Net Tend to Over-Select FeaturesLu Liu0Junheng Gao1Georgia Beasley2Sin-Ho Jung3Department of Biostatistics and Bioinformatics, Duke University, Durham, NC 27708, USADepartment of Biostatistics and Bioinformatics, Duke University, Durham, NC 27708, USADepartment of Surgery, Duke University Medical Center, Durham, NC 27710, USADepartment of Biostatistics and Bioinformatics, Duke University, Durham, NC 27708, USAMachine learning methods have been a standard approach to select features that are associated with an outcome and to build a prediction model when the number of candidate features is large. LASSO is one of the most popular approaches to this end. The LASSO approach selects features with large regression estimates, rather than based on statistical significance, that are associated with the outcome by imposing an <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msub><mi>L</mi><mn>1</mn></msub></semantics></math></inline-formula>-norm penalty to overcome the high dimensionality of the candidate features. As a result, LASSO may select insignificant features while possibly missing significant ones. Furthermore, from our experience, LASSO has been found to select too many features. By selecting features that are not associated with the outcome, we may have to spend more cost to collect and manage them in the future use of a fitted prediction model. Using the combination of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msub><mi>L</mi><mn>1</mn></msub></semantics></math></inline-formula>- and <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msub><mi>L</mi><mn>2</mn></msub></semantics></math></inline-formula>-norm penalties, elastic net (EN) tends to select even more features than LASSO. The overly selected features that are not associated with the outcome act like white noise, so that the fitted prediction model may lose prediction accuracy. In this paper, we propose to use standard regression methods, without any penalizing approach, combined with a stepwise variable selection procedure to overcome these issues. Unlike LASSO and EN, this method selects features based on statistical significance. Through extensive simulations, we show that this maximum likelihood estimation-based method selects a very small number of features while maintaining a high prediction power, whereas LASSO and EN make a large number of false selections to result in loss of prediction accuracy. Contrary to LASSO and EN, the regression methods combined with a stepwise variable selection method is a standard statistical method, so that any biostatistician can use it to analyze high-dimensional data, even without advanced bioinformatics knowledge.https://www.mdpi.com/2227-7390/11/17/3738logistic regressionmachine learningprediction modelROC curvevariable selection |
spellingShingle | Lu Liu Junheng Gao Georgia Beasley Sin-Ho Jung LASSO and Elastic Net Tend to Over-Select Features Mathematics logistic regression machine learning prediction model ROC curve variable selection |
title | LASSO and Elastic Net Tend to Over-Select Features |
title_full | LASSO and Elastic Net Tend to Over-Select Features |
title_fullStr | LASSO and Elastic Net Tend to Over-Select Features |
title_full_unstemmed | LASSO and Elastic Net Tend to Over-Select Features |
title_short | LASSO and Elastic Net Tend to Over-Select Features |
title_sort | lasso and elastic net tend to over select features |
topic | logistic regression machine learning prediction model ROC curve variable selection |
url | https://www.mdpi.com/2227-7390/11/17/3738 |
work_keys_str_mv | AT luliu lassoandelasticnettendtooverselectfeatures AT junhenggao lassoandelasticnettendtooverselectfeatures AT georgiabeasley lassoandelasticnettendtooverselectfeatures AT sinhojung lassoandelasticnettendtooverselectfeatures |