Prediction of road traffic accidents loss using improved wavelet neural network

In this paper, based on the wavelet transform theory, a wavelet network is established as an alternative to feed forward neural network for approximating arbitrary nonlinear function and an algorithm of back-propagation type is proposed for wavelet network learning. Moreover, by using principal component analysis, the major impact factors are selected, and the relationships among accident number, people death, people injury and accidents loss are systematically analyzed and modeled to predict road traffic accidents loss using wavelet neural network. The experimental results show that the wavelet network has such properties as simple structure of network, fast convergence and strong function approximation ability and provides a new prediction approach for traffic accidents loss.