A heart failure diagnosis model based on support vector machine

Abstract

To help clinicians diagnose Heart failure (HF) at the early stage, this study proposes a scoring model based on support vector machine (SVM). Missing data in clinic are imputed by employing Bayesian principal component analysis. According to the evaluation of cardiac dysfunction, samples are classified into three groups: the healthy group (without cardiac dysfunction), the HF-prone group (in asymptomatic stages of cardiac dysfunction) and the HF group (in symptomatic stages of cardiac dysfunction). The total accuracy of the model in classification is 74.4%, with accuracies of 78.79%, 87.5% and 65.85% for identifying the healthy group, the HF-prone group and the HF group, respectively. Compared with the reported results in clinical practice, the model helps to improve the accuracy of HF diagnosis,especially in screening HF patients at the early stage.