1294-GRP Optimizing Cardiovascular Risk Screening: Balancing Class Imbalance via Cost-Sensitive Learning Versus Data Resampling, 2025 Behavioral Risk Factor Surveillance System (BRFSS)

Location

https://campus.kennesaw.edu/colleges-departments/ccse/events/computing-showcase/fa26-cday-program.php

Document Type

Event

Start Date

20-11-2026 2:00 PM

Description

Predicting cardiovascular disease from public health surveys like the BRFSS is challenged by class imbalance (~9% prevalence). While SMOTE is the one accepted correction strategy, synthetic interpolation risks introducing artificial artifacts into population health data. We compare SMOTE pipelines against algorithmic cost-sensitive learning using 10-fold cross-validation and SHAP explainability. Cost-sensitive XGBoost achieved superior discrimination (ROCAUC 0.7926 vs. 0.7760) and substantially increased screening recall from 65.1% to 78.0% without expanding dataset size. Algorithmic cost-sensitive learning provides a noise-free, computationally efficient framework that substantially enhances clinical risk detection in imbalanced public health surveillance.

This document is currently not available here.

Share

COinS
 
Nov 20th, 2:00 PM

1294-GRP Optimizing Cardiovascular Risk Screening: Balancing Class Imbalance via Cost-Sensitive Learning Versus Data Resampling, 2025 Behavioral Risk Factor Surveillance System (BRFSS)

https://campus.kennesaw.edu/colleges-departments/ccse/events/computing-showcase/fa26-cday-program.php

Predicting cardiovascular disease from public health surveys like the BRFSS is challenged by class imbalance (~9% prevalence). While SMOTE is the one accepted correction strategy, synthetic interpolation risks introducing artificial artifacts into population health data. We compare SMOTE pipelines against algorithmic cost-sensitive learning using 10-fold cross-validation and SHAP explainability. Cost-sensitive XGBoost achieved superior discrimination (ROCAUC 0.7926 vs. 0.7760) and substantially increased screening recall from 65.1% to 78.0% without expanding dataset size. Algorithmic cost-sensitive learning provides a noise-free, computationally efficient framework that substantially enhances clinical risk detection in imbalanced public health surveillance.