Prediction Model for Road Traffic Accident Based on Random Forest

Abstract

More than 1.3 million people worldwide die in road traffic accidents every year, and the number of traffic accident deaths in China is increasing by 10% every year. Traffic safety has become an important direction for the development of road transportation systems. At present, research in progress of road traffic accidents is mainly from accidents numbers and the degree of accident damage. Firstly, this paper proposes a random forest prediction model for the prediction of the number of road traffic accidents, which applied to the actual to verify the validity of the model by using the R software to solve it. 1. Predictive Model It researched on the amount of the traffic accidents, which the traditional statistical models commonly included that linear regression models, poisson regression models, negative binomial distribution models, zero-inflated model and hurdle models. The following elaborated Random Forest expressions, features and in which using condition. 2. Random Forest Prediction Model The random forest method was first proposed by Leo Breiman and Adele Cutler. It is a classification algorithm that trains and analyzes samples through multiple trees. Intuitively, the decision trees which are classification models that predict categories or taking label by by inputting features. While the random forest is a model composed which multiple decision trees due to different training data. Its result is the combination of all decision tree outputs, which is more stable and with higher prediction accuracy. Taking randomly sampling with replacement in the Random forests progress, and select each tree from the total set. The relationship of the feature at this node needs to be calculated at each node, and then select one branch to move to the next node according to the operation result. We extract the use features randomly that sampling without replacement when training the nodes of each tree by Leo Breiman's proposal. This ratio may be assumed to be when S is the total number of features, then taking sqrt(S), 1/2sqrt(S), 2sqrt(S)