Detecting Congestive Heart Failure by Extracting Multimodal Features and Employing Machine Learning Techniques

Table 1

CHF detection performance based on time domain features by applying machine learning techniques.

Classifier	TPR (%)	TNR (%)	PPV (%)	NPV (%)	TA (%)	AUC	LB	UP
Decision tree (DT)
Fine	78	77	68	85	77.6	0.73	0.22	0.77
Coarse	89	66	78	81	80.2	0.75	0.11	0.66
Support vector machine (SVM)
Linear	90	73	82	84	83.6	0.92	0.10	0.73
Quadratic	88	66	76	81	79.3	0.84	0.13	0.66
Cubic	85	70	74	82	79.3	0.88	0.15	0.70
Med. Gaussian	89	73	80	84	82.8	0.90	0.11	0.73
K-nearest neighbor (KNN)
Fine	79	59	63	76	71.6	0.69	0.21	0.59
Medium	88	68	77	82	80.2	0.87	0.15	0.75
Cosine	82	80	73	87	81.0	0.83	0.18	0.80
Ensemble classifiers
Bagged tree	85	75	75	85	81.0	0.87	0.15	0.75
Subsp. disc.	96	66	91	82	84.5	0.91	0.04	0.66
RUSBoosted tree	76	68	64	80	73.3	0.81	0.24	0.68