Skip to main content

Table 3 Variable importance produced by Random Forest model trained on findings of serious misconduct

From: A machine learning analysis of serious misconduct among Australian police

Variable Mean Decrease Gini (MDG)
Demographic variables
 Secondary employment 0.13949791
 Awards or complimentary remarks 0.12657524
 Age 0.06176403
 Length of service 0.05572832
 Socio-economic status 0.03953699
 Years at rank 0.03666515
 Transfers 0.03130221
 Duty type: General duties 0.00876206
 Rank: Senior Constable 0.00698986
Misconduct process variables
 Unsubstantiated complaints 0.06648072
 Local Area Commander Warning Notice 0.05990792
 Region Commander Warning Notice 0.05457879
 Performance or conduct management plan 0.0249561
 Management action 0.02337803
Prior minor misconduct
 Issues with an investigation 0.03136233
 Improper use of force 0.02252512
 Drink driving 0.02020685
 Complaint related to court 0.01807019
 Breach code of conduct 0.01761054
 Harassment, intimidation or bullying 0.01534989
 Allegation of corruption 0.01431663
 Positive drug test 0.0108887
 Disobey reasonable direction 0.01070835
 Firearm related complaint 0.00983039
 Untruthfulness/Dishonesty 0.00948597
 Rudeness to members of the public 0.00902758
 Breach privacy, unauthorised disclosure of data 0.00840011
 Property or exhibits complaint 0.00767745