Skip to main content

Advertisement

Table 5 Mean scores on the combined MDS and RCT datasets for each of the four classifiers

From: Automatically determining cause of death from verbal autopsy narratives

  Precision Sensitivity F 1 PCCC CSMFA CCCSMFA
Adult (15–69 years)
Naïve Bayes .710 .710 .704 .689 .929 .801
Random forest .733 .730 .728 .711 .948 .854
SVM .746 .737 .740 .718 .962 .894
Neural network .773 .770 .770 .764 .962 .894
Child (29 days–14 years)
Naïve Bayes .647 .595 .608 .565 .851 .585
Random forest .687 .620 .638 .591 .872 .643
SVM .686 .658 .666 .632 .914 .760
Neural network .719 .695 .698 .672 .904 .733
Neonate (<29 days)
Naïve Bayes .507 .516 .493 .376 .826 .509
Random forest .534 .542 .524 .411 .852 .581
SVM .537 .538 .524 .404 .857 .597
Neural network .579 .576 .556 .453 .825 .507
  1. Adult and child results classified into 15 categories; neonatal records into 5 categories. Bold indicates the best score in each column for each age group. PCCC: partially chance-corrected concordance, CSMFA: cause-specific mortality fraction (CSMF) accuracy, CCCSMFA: chance-corrected CSMFA