基于改进粒子群优化算法优化LSTM-AM的公交客流量预测  

Bus Passenger Flow Forecasting Based on LSTM-AM Neural Network Optimized by Improved Particle Swarm Optimization

作　　者：连莲 穆雅伟 宗学军 何戡 商家硕 LIAN Lian;MU Yawei;ZONG Xuejun;HE Kan;SHANG Jiashuo(School of Information Engineering,Shenyang University of Chemical Technology,Shenyang 110142,China)

机构地区：[1]沈阳化工大学信息工程学院,辽宁沈阳110020

出　　处：《控制工程》2025年第2期216-225,共10页Control Engineering of China

基　　金：辽宁省“兴辽英才计划”资助项目(XLYC2002085)。

摘　　要：为提高公交客流量预测的准确性,提出一种改进的粒子群优化算法(PSO)来优化具有注意力机制的长短时记忆神经网络(LSTM-AM)的公交客流量预测模型。该模型利用注意力机制对LSTM的输入特征进行加权处理,突出对预测结果影响较大的客流特征,提高了LSTM对特征变量重要程度的提取和记忆能力。提出利用带有随机因子的非线性动态递减惯性权重并结合自适应柯西变异等操作来改进PSO的寻优性能,利用改进后的NACMPSO算法自动调整LSTM-AM模型的参数达到最优值,解决了LSTM-AM模型参数选取困难的问题,提升了客流预测精度。以公交IC卡数据和天气数据验证了该预测方法的有效性,并设置多组对比实验。结果表明,NACMPSO-LSTM-AM预测模型具有更高的预测精度和稳定性。To improve the accuracy of bus passenger flow forecasting,a bus passenger flow forecasting model based on long short-term memory neural network and attention mechanism(LSTM-AM)optimized by improved particle swarm optimization(PSO)is proposed.The model uses the attention mechanism to weight the input features of LSTM,highlighting the passenger flow features that have a greater impact on the prediction results,and improving the ability of LSTM to extract and remember the importance of feature variables.It is proposed to improve the optimization performance of PSO by using nonlinear dynamic decreasing inertia weight with random factors and adaptive Cauchy mutation.The improved NACMPSO algorithm is used to automatically adjust the parameters of LSTM-AM model to achieve the optimal value,which solves the problem of difficult parameter selection of LSTM-AM model and improves the accuracy of passenger flow forecasting.The effectiveness of the prediction method is verified by bus IC card data and weather data,and multiple sets of comparative experiments are set up.The results show that the NACMPSO-LSTM-AM prediction model has higher forecasting accuracy and stability.

关 键 词：客流量预测 粒子群优化 长短时记忆神经网络 注意力机制 公交IC卡数据 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]