«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

29

期数:

2019年12期

页码:

122-129

栏目:

应用开发研究

出版日期:

2019-12-10

文章信息/Info

Title:

Pedestrian-vehicle Model Construction and Resource Analysis System Based on Data Intelligence

文章编号:

1673-629X(2019)12-0122-08

作者:

叶 飞; 吴奇石

西南交通大学 信息科学与技术学院,四川 成都 611756

Author(s):

YE Fei; WU Qi-shi

School of Information Science and Technology,Southwest Jiaotong University,Chengdu 611756,China

关键词:

跟踪; 支持向量机; 果蝇优化; 遗传算法; 优化算法; 二值分类

Keywords:

tracking; support vector machine; fruit fly optimization; genetic algorithm; optimization algorithm; binary classification

分类号:

TP311

DOI:

10. 3969 / j. issn. 1673-629X. 2019. 12. 022

摘要:

随着生活水平的提高,越来越多的家庭开始购买不止一辆车。 庞大的需求也促进了国内汽车制造商的蓬勃发展,同时世界各品牌的汽车制造厂也纷纷进入中国市场。 然而无论是国内企业还是国外企业,产业的差异性越来越小,使得原来依赖于成本优势来提高竞争力的模式逐渐失效。 目前企业把重心放在客户体验和需求以及如何提高汽车产品的设计上。 文中以 AA 企业为原型,对其整车销售和营销模式进行分析。 目前产业链协同平台已完善了汽车制造厂从生产到入库以及汽车经销商从下销售订单计划到销售结束的业务流程。 然而,由于缺乏足够的数据分析支持,汽车制造厂和经销商往往不能估计到客户的真实需求,也不能准确地从海量的客户资源中挖掘出潜在客户。 为解决该类问题,文中提出构建基于数据智能的人车模型以及客户资源分析系统。 分析了现存的整车销售与营销模式的需求,提出一种新的混合优化算法。 该算法将遗传算法在解决离散问题上的优势和群体智能算法在解决连续问题上的优势相结合,通过寻找最优特征子集和最优支持向量机参数配置(惩罚参数和核函数参数)来优化支持向量机。 该算法的主要创新在于提出了三个并行的操作层。 其中两个层是遗传算法操作层和群体算法操作层,第三个层是协调层,主要负责接收其他两个层的个体信息并组合成新的个体信息进行评估。 随后将评估结果返回给其他两个层。 因此当算法优化 SVM 时,不需要额外的映射函数来将离散变量转化为连续变量或是连续变量转化为离散变量。 最后该算法被成功应用到产业链协同人车模型中。

Abstract:

With the improvement of living standard,more and more families begin to buy more than one car. The huge demand has also promoted the vigorous development of domestic automobile manufacturers,while the automobile manufacturers from all over the world have also entered the Chinese market. However,whether domestic or foreign enterprises,the industry difference is becoming smaller and smaller,which makes the original mode of relying on cost advantage to improve competitiveness gradually invalid. At present,enterprises are focusing on customer experience and demand and how to improve the design of automobile products. With AA enterprise as a prototype, we analyze its vehicle sales and marketing mode. At present,the industry chain collaboration platform has improved the business processes of automobile manufacturers from production to warehousing,and automobile dealers from the next sales order plan to the end of sales. However,due to the lack of sufficient data analysis support, automobile manufacturers and distributors often cannot estimate the real needs of customers. It can not accurately excavate potential customers from the vast customer resources. In order to solve these problems,we propose to build a data intelligence-based car model and customer resource analysis system. We analyze the demand of the existing vehicle sales and marketing mode,and then propose a new hybrid optimization algorithm which combines the advantages of genetic algorithm in solving discrete problems and swarm intelligence algorithm in solving continuous problems. The algorithm optimizes the support vector machine by searching for the optimal feature subset and the optimal support vector machine parameter allocation (penalty parameter and kernel function parameter). The main innovation of this algorithm is to propose three parallel operation layers. Two of them are genetic algorithm operation layer and swarm algorithm operation layer. The third layer is the coordination layer,which is mainly responsible for receiving individual information from the other two layers and combining them into new individual information for evaluation. The evaluation results are then returned to the other two layers. Therefore,when the algorithm optimizes SVM,no additional mapping function is needed to convert discrete variables into continuous variables or continuous variables into discrete variables. Finally,the algorithm is successfully applied to the collaborative human-vehicle model of industrial chain.

常用功能

更新日期/Last Update: 2019-12-10