This study uses machine learning to predict heart disease early, focusing on key risk factors to improve patient outcomes and reduce healthcare burdens. An open dataset from Kaggle, featuring data from the Behavioral Risk Factor Surveillance System (BRFSS), was used. To address the lower prevalence of heart disease, the Near Miss algorithm balanced the sample by under-sampling prevalent cases, enhancing model reliability. The study prioritizes identifying protective factors to inform preventive strategies. A Conditional Inference Tree (CIT) model was employed, with careful tuning and the Bonferroni correction ensuring statistical validity. This approach captures complex relationships and provides accurate heart disease predictions.
This report is a poster presentation of Axtria from ISPOR 2024.
Contact us at connect@axtria.com with any questions.