管理评论

›› 2020, Vol. 32 ›› Issue (3): 300-306.

• 物流与供应链管理 • 上一篇    下一篇

深度信念网络优化BP神经网络的交通流预测模型

孔繁辉1, 李健1,2   

  1. 1. 天津理工大学循环经济与企业可持续发展研究中心, 天津 300384;
    2. 天津大学管理与经济学部, 天津 300072
  • 收稿日期:2017-03-29 出版日期:2020-03-28 发布日期:2020-04-08
  • 作者简介:孔繁辉,天津理工大学循环经济与企业可持续发展研究中心博士研究生;李建,天津理工大学循环经济与企业可持续发展研究中心教授,天津大学管理与经济学部教授,博士生导师,博士
  • 基金资助:

    教育部哲学社会科学研究重大课题攻关项目（15JZD021）。

Traffic Flow Prediction Model Based on Deep Belief Network Optimized BP Neural Network

Kong Fanhui1, Li Jian1,2   

  1. 1. Research Center for Recycling Economy and Enterprises Sustainable Development, Tianjin University of Technology, Tianjin 300384;
    2. Department of Management and Economics, Tianjin University, Tianjin 300072
  • Received:2017-03-29 Online:2020-03-28 Published:2020-04-08

PDF (PC)

319

摘要：

为提高BP神经网络预测精度，基于深度学习理论提出一种深度信念网络（DBN）算法优化传统BP神经网络预测模型。该预测算法由多层限制玻尔兹曼机（RBM）组成，采用无监督学习算法训练参数，然后利用反向学习微调网络参数，进而优化BP神经网络的阈值和权值，通过训练模型求得最优解。实验表明，该预测模型克服了传统神经网络容易陷入局部最优以及函数拟合度不高的缺点，可有效提高交通流预测精度。

关键词: 交通流预测, 深度学习, 深度信念网络, BP神经网络, 限制玻尔兹曼机

Abstract:

In order to improve the prediction accuracy of BP neural network, this paper, based on deep learning theory, puts forward a deep belief network (DBN) algorithm to optimize the traditional BP neural network prediction. This algorithm is composed of multi Restricted Boltzmann Machine (RBM), and it uses unsupervised learning algorithm to train the parameters. Then, it uses reverse learning to fine tune the network parameters and optimize the threshold and weight of BP neural network. This way can derive the optimal solution through training. The experimental results show that the optimization model can overcome two shortcomings of traditional neural network:one is that traditional neural network tend to fall into local optimum and the other is that the function fitting degree remains low. Therefore, this model can effectively improve the prediction accuracy of traffic flow.

Key words: traffic flow prediction, deep learning, deep belief network, BP neural network, Restricted Boltzmann Machine