计算机工程与应用 ›› 2017, Vol. 53 ›› Issue (12): 9-15.DOI: 10.3778/j.issn.1002-8331.1702-0037

• 热点与综述 • 上一篇    下一篇

城市物流效率分析自适应DBN算法研究

李  楠   

  1. 1.中国(西安)丝绸之路研究院,西安 710100
    2.西安财经学院 管理学院,西安 710100
  • 出版日期:2017-06-15 发布日期:2017-07-04

Adaptive Deep Belief Network learning algorithm on urban logistics efficiency analysis

LI Nan   

  1. 1.China (Xi’an) Institute for Silk Road Research, Xi’an 710100, China
    2.School of Management, Xi’an University of Finance and Economics, Xi’an 710100, China
  • Online:2017-06-15 Published:2017-07-04

摘要: 深入研究了城市物流效率分析的研究现状,结合深度学习相关理论,针对具体问题构建三隐层连续型深度信念网络(DBN),对网络知识集进行了定义,提出了自适应DBN算法,分析了算法的收敛性。利用Iris数据集和Wine数据集验证了网络及算法的模式分类能力,分类精度高于双隐层深度信念网络与深度误差反向传播网络。根据新丝绸之路经济带沿线城市物流特点,以物流效率为评估目标,选取4个维度的13项指标建立评价指标体系,以20个核心节点城市为研究对象,利用自适应DBN算法和社会网络分析法(SNA)进行聚类分析,结果表明自适应DBN算法相对更为合理有效。研究结果为确定新丝绸之路经济带沿线城市物流发展策略、促进国内物流业未来的协作与发展奠定了研究基础。

关键词: 深度学习, 深度信念网络, 物流效率, 自适应, 聚类

Abstract: The paper deeply studies the researches of urban logistics efficiency, constructs a continuous deep belief network with three hidden layers (3CDBN) combining with the relevant theory of deep learning, defines the knowledge set of 3CDBN, proposes the adapted DBN algorithm and proves its convergence. And then, this paper verifies the pattern classification ability of the network and algorithm using the Iris data set and the Wine data set. The classification accuracy of the 3CDBN is better than 2CDBN and Error Back Propagation Neural Network (EBPNN). At last, according to the logistics characteristics of cities along New Silk Road, 20 core cities are chosen as research objects, adaptive DBN algorithm and Social Network Analysis (SNA) are used to carry on cluster analysis at the target of logistics efficiency evaluation, which with 4 dimensions and 13 indicators. The research results show that the adaptive DBN algorithm is more effective, which lays the foundation for determining the urban logistics strategy of cities along the New Silk Road and promoting the future cooperation and development of the domestic logistics industry.

Key words: deep learning, deep belief network, logistics efficiency, adaptive, clustering