Improved Genetic Neural Network and Its Application in Urban Traffic Environmental Impact Assessment

In this paper, we present a method that improves the standard genetic algorithms with bipolar compressed functional fitness scaling and reproduce operator based on crowding model. The improved genetic algorithm becomes a simple global search algorithm with fast convergence speed. We use it to optimize the structure and parameters of back-propagation (BP) network, which overcome the disadvantage that is easy to relapse into local minimum points and convergence speed is slow and the number of neurons in the hidden layer is determined empirically. By using the trained genetic BP network, we put forward a model that assesses the environmental impact of urban traffic. The experiment results show that the model is feasible.