Abstract

L₁ regularized logistic regression consists an important tool in data science and is dedicated to solve sparse generalized linear problems. The L₁ regularization is widely used in variable selection and estimation in generalized linear model analysis. This approach is intended to select the statistically important predictors. In this paper we compare the performance of some existing L₁ regularized logistic regression methods. The goal of our simulation study is directed toward the variable selection performance of regularized logistic regression in high dimensions. We consider three varying n (number of observations), p (number of predictors) settings and we support this comparison analysis by conducting various simulated experiments taking into consideration the correlation structure of the design matrix.

摘要

L ₁正则化逻辑回归是数据科学中的一个重要工具，致力于解决稀疏的广义线性问题。L ₁正则化广泛用于广义线性模型分析中的变量选择和估计。这种方法旨在选择统计上重要的预测变量。在本文中，我们比较了一些现有的L ₁正则化逻辑回归方法的性能。我们模拟研究的目标是针对高维正则化逻辑回归的变量选择性能。我们考虑三个不同的n（观察次数），p（预测变量的数量）设置，我们通过考虑设计矩阵的相关结构进行各种模拟实验来支持这种比较分析。