基于改进灰狼算法优化支持向量机的短期交通流预测  

Optimized SVM model for short-term traffic flow prediction based on improved gray wolf optimizer

作　　者：何祖杰 吴新烨[1] 刘中华[1] HE Zujie;WU Xinye;LIU Zhonghua(College of Architecture and Civil Engineering,Xiamen University,Xiamen 361005,China)

机构地区：[1]厦门大学建筑与土木工程学院,福建厦门361005

出　　处：《厦门大学学报（自然科学版）》2022年第2期288-297,共10页Journal of Xiamen University：Natural Science

基　　金：国家自然科学基金(11772277);福建省“2011协同创新中心”项目(2016BJC019);厦门市交通基础设施智能管养工程技术研究中心开放基金(TCIMI201803)。

摘　　要：实时、准确的短期交通流预测是智能交通系统的基础和关键技术之一.由于灰狼优化算法(GWO)存在收敛速度慢、易陷入局部最优解等缺陷,为进一步提升短期交通流预测的精度,提出了基于改进灰狼算法(IGWO)优化支持向量机(SVM)的短期交通流预测模型.首先,本文提出引入帐篷(Tent)混沌序列初始化灰狼种群,更改收敛因子的线性递减公式,对灰狼群体进化差分丰富种群多样性等方法提高算法的收敛速度和收敛精度.之后,通过对8个测试函数的计算,并与粒子群算法(PSO)、GWO进行对比,证明IGWO的先进性.最后,建立IGWO-SVM短期交通流预测模型,并通过实际数据对比分析IGWO-SVM、GWO-SVM、PSO-SVM、SVM这4种短期交通流预测模型的预测效果.对比结果表明:IGWO-SVM具有良好的鲁棒性和泛化能力,可以对短期交通流进行精确预测.Accurate short-term traffic-flow predictions are regarded as basic and key technologies of intelligent transportation system.Owing to some defects of the gray wolf optimizer,such as slow convergence speeds and the vulnerability of falling into local optimal solutions,we propose a new model based on improved grey wolf optimizer to optimize SVM to further improve the accuracy of short-term traffic-flow predictions.First,this paper introduces tent chaotic sequences to initialize the gray wolf populations,changes the linear decreasing formula of convergence factor,and enriches the population diversity to improve the convergence speed and convergence accuracy of the algorithm.Next,we compare eight test functions with the particle swarm optimization(PSO)and the grey wolf optimizer(GWO),and prove the superiority of IGWO(Improved GWO).Finally,IGWO-SVM short-term traffic-flow prediction model is established,and prediction effects of IGWO-SVM,GWO-SVM,PSO-SVM and SVM are compared and analyzed through the actual data.Results show that IGWO-SVM secures satisfactory robustness and generalization ability,and can predict short-term traffic flows accurately.

关 键 词：短期交通流预测 优化灰狼算法 Tent混沌序列 支持向量机 

分 类 号：U491.4[交通运输工程—交通运输规划与管理]