基于PSO-LSTM算法的智能交通流量预测研究  

Research on intelligent traffic flow prediction based on pso-lstm algorithm

作　　者：何芸[1] Yun He(Yunnan Jiaotong College,Kunming 650500,China)

机构地区：[1]云南交通职业技术学院

出　　处：《青海交通科技》2019年第4期76-79,共4页Qinghai Transportation Science and Technology

摘　　要：交通流量本身具有很强的不确定性,复杂、多变,容易受到随机因素的扰动,并且规律性不明显。随着不同预测方法的引入,对短时交通流的预测也出现了许多预测模型。长短期记忆模型擅长于处理和预测时间序列中具有时间间隔和延迟相对较长特征的重要事件。粒子群算法是一种通过模拟鸟群捕食行为设计的随机优化技术。本研究引入粒子群算法优化长短期记忆模型,使用昆明市某个路口断面所采集的过车数据作为训练集和测试集。本研究使用matlab软件进行上述模型的建模和预测,使用均方误差模型进行预测模型的误差分析。结果表明,引入粒子群算法优化模型后,预测误差降低60%,PSO算法优化LSTM模型能够更为准确的预测交通流量。Traffic flow itself has strong uncertainty,complex,changeable,easy to be disturbed by random factors,and the regularity is not obvious.With the introduction of different forecasting methods,there are many forecasting models for short-time traffic flow.The long and short term memory model is good at processing and predicting important events with relatively long time interval and delay in time series.Particle swarm optimization(pso)is a stochastic optimization technique that simulates the design of predation behavior of birds.In this study,particle swarm optimization(pso)algorithm was introduced to optimize the long and short term memory model.In this study,matlab software was used for modeling and prediction of the above models,and the mean square error model was used for error analysis of the prediction model.The results show that the prediction error is reduced by 60%after particle swarm optimization model is introduced,and the LSTM model optimized by PSO algorithm can predict traffic flow more accurately.

关 键 词：交通流量 PSO LSTM 预测 

分 类 号：U491.1[交通运输工程—交通运输规划与管理]