Skip to main content

Table 7 Performances of machine learning-based methods

From: Discovering and identifying New York heart association classification from electronic health records

NYHA Classification Precision Recall F-Measure
Feature Set 1: bag-of-words
Support Vector Machine
 I 84.71% 82.06% 83.36%
 II 88.30% 91.80% 90.01%
 III 88.34% 88.46% 88.40%
 IV 80.00% 70.81% 75.13%
 Overall 85.34% 83.28% 84.23%
Logistic Regression
 I 83.61% 79.97% 81.75%
 II 86.87% 91.99% 89.36%
 III 87.66% 88.46% 88.06%
 IV 80.72% 64.11% 71.47%
 Overall 84.72% 81.13% 82.66%
Random Forest
 I 87.54% 86.93% 87.24%
 II 91.23% 94.40% 92.79%
 III 91.83% 90.40% 91.11%
 IV 79.89% 72.25% 75.88%
 Overall 87.63% 86.00% 86.75%
Feature Set 2: n-gram
Support Vector Machine
 I 95.05% 93.73% 94.39%
 II 95.71% 96.81% 96.26%
 III 94.95% 95.20% 95.08%
 IV 89.66% 87.08% 88.35%
 Overall 93.84% 93.21% 93.52%
Logistic Regression
 I 93.09% 91.46% 92.27%
 II 94.74% 95.66% 95.20%
 III 93.12% 94.81% 93.96%
 IV 87.18% 81.34% 84.16%
 Overall 90.03% 90.82% 90.42%
Random Forest
 I 97.02% 96.52% 96.77%
 II 97.58% 97.49% 97.54%
 III 93.01% 96.63% 94.78%
 IV 93.99% 82.30% 87.76%
 Overall 95.40% 92.23% 93.78%