What question did this study set out to answer?

This research investigates how linear classifier models can optimize binary classification using a custom loss function.

April 18, 20261 citations

Linear Classifier Models for Binary Classification

HJHimchan JeongSimon Fraser University BZBin ZouToronto Metropolitan University

Key Points

This research investigates how linear classifier models can optimize binary classification using a custom loss function.
Developed a flexible loss function with penalty terms for false positives and negatives.
Conducted a simulated study and empirical analysis to evaluate model performance.
Tested various parametric weight functions for effective classification.
Linear classifiers demonstrated superior performance compared to logistic regression.
Results indicated that specific weight functions allowed targeted control over false positive and negative rates.

Abstract

We apply a class of linear classifier models under a flexible loss function to study binary classification problems. The loss function consists of two penalty terms—one penalizing false positive (FP) and the other penalizing false negative (FN)—and can accommodate various classification targets by choosing a weighting function to adjust the impact of FP and FN on classification. We show, through both a simulated study and an empirical analysis, that the linear classifier models under certain parametric weight functions can outperform the logistic regression model and can be trained to meet flexible targeted rates on FP or FN. This work was supported by a 2022 Individual Research Grant from the CAS.

Ask AI

Helpful

Bookmark

View Full Paper