A filter approach for feature selection in classification: application to automatic atrial fibrillation detection in electrocardiogram recordings

Michel, Pierre; Ngo, Nicolas; Pons, Jean-François; Delliaux, Stéphane; Giorgi, Roch

doi:10.1186/s12911-021-01427-8

BMC Medical Informatics and Decision Making

Table 4 Classification performances (median accuracy, sensitivity, specificity, and MCC values with their interquartile ranges) computed by bootstrap with 1000 replications applied to the validation dataset

From: A filter approach for feature selection in classification: application to automatic atrial fibrillation detection in electrocardiogram recordings

Features		Classification performances (Validation dataset)
Dim	Added feature	Accuracy	Specificity	Sensitivity	MCC
\(\gamma\)-metric
1	\(sd_2\)	\(0.9108\ (0.0019)\)	\(0.9518\ (0.0016)\)	\(0.7983\ (0.0054)\)	\(0.768\ (0.0051)\)
2	\(sd_3\)	\(0.8973\ (0.0020)\)	\(0.9502\ (0.0017)\)	\(0.7520\ (0.0051)\)	\(0.730\ (0.0053)\)
3	\(sd_1\)	\(0.8972\ (0.0019)\)	\(0.9502\ (0.0017)\)	\(0.7518\ (0.0054)\)	\(0.730\ (0.0052)\)
4	pNN50	\(0.8945\ (0.0022)\)	\(0.9449\ (0.0018)\)	\(0.7565\ (0.0055)\)	\(0.724\ (0.0054)\)
5	\(sd_4\)	\(0.8945\ (0.0020)\)	\(0.9448\ (0.0017)\)	\(0.7563\ (0.0054)\)	\(0.724\ (0.0051)\)
6	RMSSD	\(0.8933\ (0.0020)\)	\(0.9445\ (0.0018)\)	\(0.7531\ (0.0051)\)	\(0.721\ (0.0052)\)
7	SDSD	\(0.8935\ (0.0021)\)	\(0.9446\ (0.0018)\)	\(0.7533\ (0.0058)\)	\(0.721\ (0.0054)\)
8	\(m_2\)	\(0.8944\ (0.0021)\)	\(0.9456\ (0.0018)\)	\(0.7543\ (0.0057)\)	\(0.724\ (0.0053)\)
9	\(m_3\)	\(0.8941\ (0.0020)\)	\(0.9464\ (0.0018)\)	\(0.7509\ (0.0057)\)	\(0.722\ (0.0052)\)
10	\(m_1\)	\(0.8931\ (0.0021)\)	\(0.9448\ (0.0018)\)	\(0.7512\ (0.0056)\)	\(0.720\ (0.0052)\)
11	SDNNIDX	\(0.8983\ (0.0021)\)	\(0.9492\ (0.0016)\)	\(0.7586\ (0.0056)\)	\(0.733\ (0.0054)\)
12	\(sd_5\)	\(0.8980\ (0.0021)\)	\(0.9500\ (0.0017)\)	\(0.7552\ (0.0057)\)	\(0.732\ (0.0052)\)
13	\(m_4\)	\(0.8980\ (0.0019)\)	\(0.9490\ (0.0016)\)	\(0.7583\ (0.0058)\)	\(0.733\ (0.0050)\)
14	MADDR	\(0.8961\ (0.0020)\)	\(0.9474\ (0.0016)\)	\(0.7551\ (0.0056)\)	\(0.728\ (0.0054)\)
15	\(sd_6\)	\(0.8957\ (0.0021)\)	\(0.9474\ (0.0018)\)	\(0.7542\ (0.0052)\)	\(0.727\ (0.0054)\)
16	\(m_5\)	\(0.8955\ (0.0021)\)	\(0.9465\ (0.0018)\)	\(0.7554\ (0.0055)\)	\(0.726\ (0.0054)\)
Mean decrease accuracy
1	\(sd_3\)	\(0.8930\ (0.0022)\)	\(0.9517\ (0.0017)\)	\(0.7317\ (0.0059)\)	\(0.718\ (0.0060)\)
2	\(sd_2\)	\(0.8972\ (0.0021)\)	\(0.9502\ (0.0017)\)	\(0.7517\ (0.0051)\)	\(0.730\ (0.0055)\)
3	\(sd_4\)	\(0.8972\ (0.0019)\)	\(0.9506\ (0.0017)\)	\(0.7509\ (0.0053)\)	\(0.730\ (0.0053)\)
4	\(m_3\)	\(0.8975\ (0.0020)\)	\(0.9514\ (0.0014)\)	\(0.7498\ (0.0055)\)	\(0.731\ (0.0051)\)
5	\(sd_5\)	\(0.8970\ (0.0020)\)	\(0.9513\ (0.0017)\)	\(0.7484\ (0.0054)\)	\(0.729\ (0.0053)\)
6	\(m_2\)	\(0.8979\ (0.0019)\)	\(0.9512\ (0.0017)\)	\(0.7520\ (0.0052)\)	\(0.732\ (0.0051)\)
7	\(sd_6\)	\(0.8977\ (0.0020)\)	\(0.9514\ (0.0016)\)	\(0.7506\ (0.0059)\)	\(0.731\ (0.0052)\)
8	\(m_4\)	\(0.8971\ (0.0021)\)	\(0.9504\ (0.0017)\)	\(0.7510\ (0.0057)\)	\(0.730\ (0.0053)\)
9	\(sd_7\)	\(0.8976\ (0.0019)\)	\(0.9512\ (0.0017)\)	\(0.7505\ (0.0057)\)	\(0.731\ (0.0051)\)
10	\(sd_1\)	\(0.8978\ (0.0020)\)	\(0.9511\ (0.0016)\)	\(0.7516\ (0.0059)\)	\(0.731\ (0.0051)\)
11	\(sd_8\)	\(0.8981\ (0.0021)\)	\(0.9515\ (0.0016)\)	\(0.7516\ (0.0060)\)	\(0.732\ (0.0055)\)
12	\(sd_9\)	\(0.8977\ (0.0021)\)	\(0.9522\ (0.0016)\)	\(0.7482\ (0.0057)\)	\(0.731\ (0.0053)\)
13	\(m_1\)	\(0.8972\ (0.0020)\)	\(0.9510\ (0.0016)\)	\(0.7499\ (0.0058)\)	\(0.730\ (0.0052)\)
14	\(sd_{10}\)	\(0.8970\ (0.0020)\)	\(0.9514\ (0.0016)\)	\(0.7475\ (0.0055)\)	\(0.729\ (0.0054)\)
15	\(m_5\)	\(0.8978\ (0.0021)\)	\(0.9522\ (0.0016)\)	\(0.7487\ (0.0055)\)	\(0.732\ (0.0054)\)
16	SDSD	\(0.8989\ (0.0020)\)	\(0.9532\ (0.0017)\)	\(0.7499\ (0.0054)\)	\(0.734\ (0.0053)\)
Mean decrease gini
1	\(sd_3\)	\(0.8931\ (0.0022)\)	\(0.9516\ (0.0017)\)	\(0.7321\ (0.0058)\)	\(0.718\ (0.0056)\)
2	\(sd_2\)	\(0.8973\ (0.0021)\)	\(0.9503\ (0.0017)\)	\(0.7517\ (0.0051)\)	\(0.730\ (0.0054)\)
3	\(m_3\)	\(0.8977\ (0.0019)\)	\(0.9509\ (0.0016)\)	\(0.7517\ (0.0054)\)	\(0.731\ (0.0051)\)
4	\(m_4\)	\(0.8972\ (0.0021)\)	\(0.9506\ (0.0017)\)	\(0.7512\ (0.0056)\)	\(0.730\ (0.0053)\)
5	\(sd_4\)	\(0.8973\ (0.0019)\)	\(0.9507\ (0.0016)\)	\(0.7507\ (0.0052)\)	\(0.730\ (0.0051)\)
6	\(m_2\)	\(0.8983\ (0.0019)\)	\(0.9509\ (0.0016)\)	\(0.7541\ (0.0050)\)	\(0.733\ (0.0051)\)
7	\(m_5\)	\(0.8975\ (0.0020)\)	\(0.9510\ (0.0017)\)	\(0.7507\ (0.0060)\)	\(0.731\ (0.0052)\)
8	\(sd_5\)	\(0.8977\ (0.0020)\)	\(0.9510\ (0.0017)\)	\(0.7515\ (0.0057)\)	\(0.731\ (0.0053)\)
9	\(sd_1\)	\(0.8975\ (0.0020)\)	\(0.9504\ (0.0017)\)	\(0.7528\ (0.0056)\)	\(0.731\ (0.0051)\)
10	\(sd_6\)	\(0.8974\ (0.0020)\)	\(0.9513\ (0.0016)\)	\(0.7497\ (0.0059)\)	\(0.731\ (0.0052)\)
11	\(m_1\)	\(0.8964\ (0.0020)\)	\(0.9504\ (0.0016)\)	\(0.7486\ (0.0060)\)	\(0.728\ (0.0055)\)
12	\(sd_7\)	\(0.8970\ (0.0021)\)	\(0.9513\ (0.0017)\)	\(0.7478\ (0.0060)\)	\(0.729\ (0.0054)\)
13	\(m_6\)	\(0.8974\ (0.0020)\)	\(0.9529\ (0.0015)\)	\(0.7456\ (0.0060)\)	\(0.730\ (0.0052)\)
14	\(sd_8\)	\(0.8977\ (0.0021)\)	\(0.9531\ (0.0015)\)	\(0.7459\ (0.0056)\)	\(0.731\ (0.0053)\)
15	\(sd_9\)	\(0.8978\ (0.0021)\)	\(0.9537\ (0.0016)\)	\(0.7447\ (0.0056)\)	\(0.731\ (0.0055)\)
16	\(sd_{10}\)	\(0.8981\ (0.0020)\)	\(0.9536\ (0.0017)\)	\(0.7459\ (0.0055)\)	\(0.732\ (0.0052)\)
AUC
1	\(sd_3\)	\(0.8930\ (0.0022)\)	\(0.9517\ (0.0017)\)	\(0.7317\ (0.0059)\)	\(0.718\ (0.0056)\)
2	\(sd_2\)	\(0.8973\ (0.0020)\)	\(0.9502\ (0.0017)\)	\(0.7317\ (0.0059)\)	\(0.730\ (0.0053)\)
3	\(m_3\)	\(0.8977\ (0.0019)\)	\(0.9509\ (0.0016)\)	\(0.7516\ (0.0055)\)	\(0.731\ (0.0051)\)
4	\(sd_4\)	\(0.8975\ (0.0020)\)	\(0.9514\ (0.0017)\)	\(0.7501\ (0.0056)\)	\(0.731\ (0.0053)\)
5	\(m_2\)	\(0.8974\ (0.0020)\)	\(0.9515\ (0.0016)\)	\(0.7490\ (0.0054)\)	\(0.730\ (0.0052)\)
6	\(sd_5\)	\(0.8977\ (0.0019)\)	\(0.9515\ (0.0017)\)	\(0.7502\ (0.0051)\)	\(0.731\ (0.0052)\)
7	\(m_4\)	\(0.8976\ (0.0020)\)	\(0.9507\ (0.0016)\)	\(0.7520\ (0.0059)\)	\(0.731\ (0.0052)\)
8	\(sd_1\)	\(0.8974\ (0.0021)\)	\(0.9506\ (0.0017)\)	\(0.7518\ (0.0056)\)	\(0.731\ (0.0053)\)
9	\(sd_6\)	\(0.8971\ (0.0019)\)	\(0.9504\ (0.0017)\)	\(0.7512\ (0.0057)\)	\(0.730\ (0.0052)\)
10	\(m_1\)	\(0.8965\ (0.0020)\)	\(0.9508\ (0.0016)\)	\(0.7473\ (0.0060)\)	\(0.728\ (0.0051)\)
11	\(sd_7\)	\(0.8967\ (0.0021)\)	\(0.9504\ (0.0016)\)	\(0.7496\ (0.0059)\)	\(0.729\ (0.0054)\)
12	\(m_5\)	\(0.8970\ (0.0021)\)	\(0.9513\ (0.0017)\)	\(0.7478\ (0.0060)\)	\(0.729\ (0.0054)\)
13	\(sd_8\)	\(0.8971\ (0.0020)\)	\(0.9519\ (0.0015)\)	\(0.7470\ (0.0057)\)	\(0.730\ (0.0052)\)
14	\(sd_9\)	\(0.8974\ (0.0020)\)	\(0.9525\ (0.0016)\)	\(0.7464\ (0.0055)\)	\(0.730\ (0.0054)\)
15	\(sd_{10}\)	\(0.8978\ (0.0021)\)	\(0.9522\ (0.0016)\)	\(0.7486\ (0.0055)\)	\(0.731\ (0.0054)\)
16	RMSSD	\(0.8991\ (0.0020)\)	\(0.9526\ (0.0018)\)	\(0.7522\ (0.0055)\)	\(0.735\ (0.0051)\)

Results are presented according to the number of features (from 1 to 16) selected by each feature selection approach (\(\gamma\)-metric, MDA, MDG, and AUC). Interquartile ranges were computed on the bootstrap estimations of the performance indicators as a measure of dispersion

Back to article page

ISSN: 1472-6947

Contact us

General enquiries: journalsubmissions@springernature.com