Table 1.
Data summary.
Fig 1.
Hyperparameter tuning in KNN.
Table 2.
Performance of classification models for crash severity levels.
Fig 2.
Comparison performance measures of classifiers for different VRU models.
Fig 3.
ROC curves of different classifiers.
(A) Motorcyclists (B) Bicyclists (C) Pedestrians (D) VRUs.
Fig 4.
Random forest based feature ranking for QLD VRU.
Fig 5.
Partial dependency plots of different features with respect to VRU crash severity.
(A) Age Group (B) Disobey Road Rule (C) Drink, Drug and Alcohol Related (D) Area Remoteness (E) Day of Week (F) Driver Condition.
Fig 6.
Partial dependency plots of different features with respect to VRU crash severity.
(G) Roadway Features (H) Hour (Time of Day) (I) Road Section Authority (J) Road Region (K) Month (L) Road and Environment Condition.
Fig 7.
Partial dependency plots of different features with respect to VRU crash severity.
(M) Year (N) Speed Limit (O) Vehicle Condition (P) Traffic Control (Q) Speed Driving.