Prediction of Traffic Accidents using Random Forest Model

With the increasing trend of the accident rate, the number of casualties in humans has increased considerably over the past decades, which has led to the use of cameras, or fixed speed cameras to carry out their routine activities. In this paper, we focus on severity prediction of traffic accidents, which is a huge step in road accident management in the road. This problem provides important information for emergency logistical transportation in many cities. To evaluate the severity of road accidents in the crowded target, we evaluate the potential impact of the accident to realize effective accident management procedures. In this proposed study, we implement and compare some algorithms in machine learning such as Random Forest, Support Vector Machine, and Artificial Neural Network to classify and predict severity for Traffic accidents, and we presented the confusion matrix to specify the impact of different classes on each other for: Pedestrian, Vehicle or pillion passenger, or Driver or rider to validate this experimentation. In the numerical example we use the TRAFFIC ACCIDENTS_2019_LEEDS data from the Road Safety of department Transport to classify the Severity prediction for Traffic accidents into three classes: Pedestrian, vehicle or pillion passenger, and driver or rider to have a 93% accuracy for Random Forest compared to 82% for SVM and 87 for ANN, and at the level of precision recall we also have 93.82% for Random Forest compared to 82.22% for SVM and 87.88% for ANN.