Abstract: Aiming at the problem of rail transit passenger flow short-term forecasting, a prediction model of Wavelet Neural Network（WNN） optimized by Bat Algorithm（BA） based on Adaptive t Distribution Mutation（ATM-BA） is proposed. To make the BA can jump out of the premature convergence, the adaptive t distribution mutation with the linear decreasing control factor is introduced into the BA. And then, the ATM-BA and WNN are combined to make use of the ATM-BA to do the optimization of the parameter configuration of the WNN which can improve the prediction accuracy of the WNN.It utilizes the ATM-BA-WNN model to forecast the short-term passenger flow of the Zhengzhou Metro Line 1, and compares with the conventional WNN prediction model, WNN optimized by BA（BA-WNN） prediction model and Support Vector Machine（SVM） prediction model. The simulation results show that the proposed model which has higher prediction accuracy, better fitting capability and smaller error, etc is more feasible and more superior by compared with other three models.

Key words: short-term passenger flow, bat algorithm, wavelet neural network, parameter optimization, short-term prediction

摘要： 针对城市轨道交通短时客流量预测问题，提出了一种基于自适应[t]分布变异的蝙蝠算法（ATM-BA）优化的小波神经网络（WNN）预测模型（ATM-BA-WNN）。在基本蝙蝠算法（BA）中引入带有线性递减控制因子的自适应[t]分布变异，使其具有变异机制，能够跳出早熟收敛。并将ATM-BA与WNN两者相互耦合，利用ATM-BA优化WNN的参数配置，进而提高WNN的预测精度。运用ATM-BA-WNN模型对郑州地铁1号线短时客流量进行预测，并与传统的WNN预测模型、BA优化的WNN（BA-WNN）预测模型以及支持向量机（SVM）预测模型进行比较。仿真结果表明，相较于其他3种模型，所建预测模型预测精度最高，拟合能力更强，误差最小，从而证明了该模型在短时客流量预测领域的可行性及优越性。

关键词: 短时客流量, 蝙蝠算法, 小波神经网络, 参数优化, 短时预测