Performance Classification for Lasso Weights with Penalized Logistic Regression for High-Dimensional Data

Abstract

In high-dimensional data, classification performance is a crucial consideration. One method of interest is the penalized binary logistic regression. However, the Least Absolute Shrinkage and Selection Operator) The Lasso method may face problems when the appropriate penalty for each coefficient is not determined. For this reason, different weights are used in weighted Lasso estimates to address this issue and improve classification performance. To overcome this limitation, we employ various Weighted Lasso Estimates, each with unique weight assignments, and compare their performance with our fifth proposed weight configuration. This application of Lasso weighting schemes aims to uncover the most effective approach for high-dimensional classification tasks while considering the optimal set of variables.

The evaluation criteria for these methods include the number of selected variables, classification accuracy, and mean squared error. We then apply these techniques to real-world data to identify the most effective classification mode and select the optimal set of variables. This rigorous and precise investigation aims to provide a robust and reliable classification approach for high-dimensional systems.

Paper type: Research paper.