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

Table 4

CHF detection performance based on entropy-based features by applying machine learning techniques.

Classifier	TPR (%)	TNR (%)	PPV (%)	NPV (%)	TA (%)	AUC	LB	UP
Decision tree (DT)
Fine	71	50	51	70	62.9	0.65	0.29	0.50
Coarse	90	36	70	70	69.8	0.69	0.10	0.36
Support vector machine (SVM)
Linear	93	30	72	68	69.0	0.71	0.07	0.30
Quadratic	83	57	68	76	73.3	0.74	0.17	0.57
Cubic	82	52	64	74	70.7	0.73	0.18	0.52
Med. Gaussian	94	30	76	69	69.8	0.75	0.06	0.30
K-nearest neighbor (KNN)
Fine	75	57	58	74	68.1	0.66	0.25	0.57
Medium	85	50	67	73	71.6	0.69	0.15	0.50
Cosine	82	52	64	74	70.7	0.72	0.18	0.52
Ensemble classifiers
Bagged tree	82	57	66	76	72.4	0.78	0.18	0.57
Subsp. disc.	89	39	68	70	69.8	0.71	0.11	0.39
RUSBoosted tree	75	59	59	75	69.0	0.75	0.25	0.59