Research on Short-term Traffic Flow Prediction Based on Wavelet Neural Network with Improved Genetic Algorithm

Abstract

Due to the time-varying and uncertainty of traffic flow, the Intelligent Traffic System (ITS) is applied more and more widely in many cities. Accurate short-term traffic flow forecasting is the core component of the system. An improved Wavelet Neural Network (WNN) prediction model based on the Genetic Algorithm (GA) is put forward in this paper. The Genetic Algorithm proposed according to the law of biological evolution is introduced, and the crossover and mutation probability of the algorithm is improved in this paper. The stretching scale and translation scale of the wavelet function are trained firstly, and then the parameters obtained after the pre-trained are further optimized by adopting the gradient descent method. It overcomes the limitation that the traditional WNN is apt to get the local minimum when optimizing parameters. The simulation results indicate that the WNN model based on the improved Genetic Algorithm (IGA) can forecast the short-term traffic flow more precisely.