Skip to main content

Table 4 Model selection for outcome variable laytime by quartile

From: Factors and predictors of length of stay in offenders diagnosed with schizophrenia - a machine-learning-based approach

Variable

Best algorithm/ statistical procedure

Accuarcy (%)

AUC

Sensivity (%)

Specificity (%)

PPV (%)

NPV (%)

Victim injured severly/fatally

Bagged trees

73.6

0.64

70

79

83

64

Index crime: (attempted) homicide

KNN

65.3

0.59

62

70

78

53

Victim injury severly/fatally and index crime (attempted) homicide

SVM

73.6

0.71

78

79

75

55