Journal of Transportation Systems Engineering and Information Technology

Journal of Transportation Systems Engineering and Information Technology ›› 2023, Vol. 23 ›› Issue (5): 75-82.DOI: 10.16097/j.cnki.1009-6744.2023.05.008

Previous Articles     Next Articles

Short-term Traffic Flow Prediction Based on ASO-ELM Hybrid Optimization Model

CAI Hao1,LI Lin-feng1,LI Han1,LI Xin*1,ZHOU Teng2   

  1. 1. Department of Computer Science and Technology, Shantou University, Shantou 515063, Guangdong, China; 2. School of Cyberspace Security, Hainan University, Haikou 570228, China
  • Received:2023-05-08 Revised:2023-07-05 Accepted:2023-08-31 Online:2023-10-25 Published:2023-10-22
  • Supported by:
    2021 Guangdong Province Special Fund for Science and Technology (STKJ2021021)

基于极限学习机的短期交通流预测混合优化模型

蔡浩1，李林峰1，李涵1，李新*1，周腾2   

  1. 1. 汕头大学，计算机科学与技术系，广东 汕头 515063；2. 海南大学，网络空间安全学院，海口 570228
  • 作者简介:蔡浩(1974- )，男，湖南岳阳人，教授，博士。
  • 基金资助:
    2021广东省科技重点专项(STKJ2021021)

PDF

Abstract: Due to the dynamic, uncertain and nonlinear characteristics of the short-term traffic flow, it is difficult to predict traffic flow accurately. In this paper, we build an ASO-ELM short-term traffic flow prediction hybrid optimization model based on Extreme Learning Machine (ELM) by embedding Atom Search Optimization (ASO). The hybrid optimization model is used to explore the prediction performance of the hybrid optimization model in the field of short-term traffic flow prediction by comparing the existing short-term traffic flow prediction models. The A10 ring road in Amsterdam, the Netherlands, is selected as the prototype of the road network, and the ASO-ELM hybrid model is used to compare with common traffic flow forecasting models for simulation forecasting experiments. The experimental results show that the mean absolute percentage error (MAPE) of the ASO-ELM hybrid model decreases by 4.3%, 3.5%, 6.9% and 5.4%, respectively, and the root mean squared error (RMSE) decreases by 4.8%、4.0%、2.0% and 5.2%, respectively. Secondly, MAPE decreased by 9.6%, 8.6%, 9.8% and 5.0%, respectively, and RMSE decreased by 4.5%, 5.9%, 2.6% and 1.7%, respectively, compared to the Artificial Neural Network (ANN). The research results reveal the potential of hybrid optimization models in the field of short-term traffic flow forecasting and provide an important basis for model exploration in the field of short-term traffic flow forecasting.

Key words: intelligent transportation, short-term traffic flow forecasting, hybrid forecasting models, atom search optimization, extreme learning machine

摘要： 交通流的动态性、不确定性和非线性等特性导致交通流难以精确预测，本文在极限学习机(Extreme Learning Machine, ELM)的基础上，通过嵌入原子搜索算法(Atom Search Optimization,ASO)，构建ASO-ELM短期交通流预测混合优化模型，对比现有短期交通流预测模型，分析混合优化模型在短期交通流预测领域的表现。实验选取荷兰阿姆斯特丹市A10环形公路为路网原型，使用ASO-ELM混合模型与常见交通流预测模型进行对比实验。实验结果表明：ASO-ELM混合模型在 4 个数据集下的平均绝对百分比误差(Mean Absolute Percentage Error, MAPE)相较于ELM模型分别下降了4.3%、3.5%、6.9%和5.4%，均方根误差(Root Mean Squared Error, RMSE)分别下降了4.8%、4.0%、2.0%和5.2%；其次，与人工神经网络(Artificial Neural Network, ANN)相比，MAPE分别下降了9.6%、8.6%、9.8%和5.0%，RMSE也分别下降了4.5%、5.9%、2.6%和1.7%。研究成果揭示了混合优化模型在短期交通流预测领域的潜力。

关键词: 智能交通, 短期交通流预测, 混合预测模型, 原子搜索算法, 极限学习机

CLC Number: