【作者】 范黎林；

【导师】 孙林夫；

【作者基本信息】 西南交通大学 ， 计算机应用技术， 2009， 博士

【摘要】 汽车产业链是由整车制造商、供应商、经销商、服务商等相关企业组成的庞大网络。基于ASP(Application Service Provider, ASP)模式的产业链协作平台,是降低产业链的运营成本、实现产业链协作中复杂信息交互的有效途径,已成为提升汽车制造业竞争能力的主要手段之一。这种公共服务平台的成功实施,提高了汽车产业链上下游企业的协作效率,降低了企业信息交换、传递和获取的成本,促进了企业业务数据的积累。随着企业业务数据的积累越来越多,以及市场竞争环境的不断加剧,如何利用这些数据创造价值成为企业下一步思考的问题。企业希望能够把大量的业务数据迅速转换成可靠的信息,发现数据背后隐藏的知识,找到潜在的规律,从而提高决策质量,把握和发现市场机遇,提升企业的竞争力。但是,传统的信息管理系统已经难以满足企业的这些新需求,由数据仓库、联机分析处理和数据挖掘等技术支撑的商务智能(Business Intelligence, BI)给企业带来了新希望。BI是一种基于大量数据基础上的提炼和重新整合知识的过程,这个过程与知识共享和知识创造紧密结合,完成了从信息到知识的转变,帮助企业决策者做出及时、正确、可行和有效的决策,最终增强企业的竞争优势。本文以四川省制造业信息化研究院开发的汽车产业链协作ASP平台为基础,依托国家科技支撑计划课题“面向高竞争性行业的产业链协作技术集成推广应用(2006BAF01A37)"的支持,进行了以下几个方面的研究工作：1)随着汽车产业链协作ASP平台的成功推广实施,企业数据的积累量与日俱增。面对加剧的市场竞争环境,企业管理者考虑在现有基础上实施商务智能系统,以满足在决策信息方面的需求。本文结合汽车产业链平台的特点以及用户的需求,提出了基于汽车产业链协作平台的商务智能系统的解决方案和体系架构；2)车辆故障分析是挖掘故障配件与车辆行驶里程、使用时间、使用地区、配件品牌、相关故障件等信息之间的关联规则,其目的是为了提高整车设计的合理性。关联规则挖掘的核心算法是寻找频繁集,经典的Apriori关联规则挖掘算法采用的候选项集产生原理决定了该算法寻找频繁集时对数据库扫描过于频繁；无候选项集FP-Growth算法,通过FP-tree寻找频繁集,极大地减少了I／O数据交换,但是,FP-tree的规模和数据特征有关,当数据集大且过于稀疏时,构造基于内存的FP-tree是不现实的。而且由于先天性“剪枝”算法的不足,导致这两种算法在数据库发生变化时,原来发现的频繁集很难重用。针对以上问题,提出了基于矩阵的频繁集发现及更新算法；3)客户是企业的重要战略资源,高效的客户关系管理以扎实的客户细分为基础,而实现客户细分的主要途径就是客户聚类。设计高效、准确的客户细分聚类算法,成为汽车协作产业链平台实施客户细分的关键。针对经典的FKP混合数据聚类算法存在收敛速度慢、初始聚类中心选择困难等问题,提出一种利用遗传算法快速搜索FKP算法需要的初始聚类中心、同时利用FKP算法对染色体进行优化、避免早熟收敛的GA-FKP聚类算法。GA_FKP聚类挖掘算法是对混合数据聚类算法改进,具有一定的通用性；基于频繁矩阵的频集发现算法,从一个全新的角度,发现频繁集,弥补了现有算法的不足；基于汽车产业链协作平台的商务智能系统架构实现了和原有平台系统的无缝结合以及程序资源共享,最大限度减少了系统开发工作量,为其他中小企业的信息系统集成BI功能提供一个参考方法。更多还原

【Abstract】 The automobile industrial chain is a huge network, which is composed of automobile manufacturers, suppliers, dealers and service providers etc. ASP-based industrial chain collaborated platform is an effective way to cut the operation cost of industrial chain, and realize the complicated information exchange in industrial chain collaboration, and the platform has become a major way to enhance the competitiveness of automobile industry. The successful implementation of the public technology service platform improves the collaborative efficiency of upstream and downstream business in automobile industrial chain, lowers the cost of business information exchange, transmission and acquisition, and accelerates the accumulation of business data.With the accumulation of enterprise’s business data and competition of market environment, how to make use of these data to create value becomes the next issue needed to be considered by the enterprises. The enterprises hope to transform these data into reliable information, find the potential rule, improve the quality of decision, grasp and discover the market opportunity, and enhance the enterprise competitiveness. But the traditional information system can not meet the new requirement of enterprise. However, the business intelligence based on data warehouse, online analytical processing and data mine brings new hope for enterprise. Business intelligence is a process based on the extraction of mass data and knowledge re-integration, this process combines closely with knowledge share and knowledge creation, accomplishes the transformation from information to knowledge, and helps the enterprise policy maker make a timely, correct, feasible and effective decision, and finally strengthens the enterprise’s competitiveness.Based on automobile industrial chain collaborative ASP platform developed by Sichuan Institute of Manufacturing Information, which is supported by National technology project " highly competitive business-oriented promotion and application of industrial chain collaborative technology integration (2006BAF01A37)", following aspects of research work are finished.1) With the successful promotion and implementation of automobile industrial collaborative ASP Platform, The accumulation of enterprise data has increased day by day. Face to the increased competition in the market environment, the managers of enterprise consider to implement the BI system based on current situation, to meet the requirements of information decision. Combined with the feature of automobile industrial chain and customer requirements, the solution and architecture of BI system based on automobile industrial chain collaborative platform is put forward.2) Faulty analysis of vehicles is to find the association rules between faulty accessories and vehicle mileage, time, region, brand of accessory, related faulty accessory, which is to improve the performance of the overall design. The core algorithm of association rules mining is to find frequent item set. The candidate item set generation principle, used by the classical association rules mining algorithm Apriori, determines that scanning the database of this algorithm is too frequently when finding the frequent item set. FP-Growth algorithm of none candidate item set reduces the I/O data exchange greatly through finding frequent set by FP-tree. But the scale of FP-tree is related to data characteristic. Constructing FP-tree based on the memory is not realistic when data set is big and too sparse. And the congenital deficiency of the pruning algorithm causes that the frequent set is difficult to reuse when the database changes. For the issue mentioned above, frequent set discovering and updating algorithm based on matrix are put forward.3) The customers are the important strategic resources of enterprises; efficient customer relation management is on the basis of solid customer segmentation. And the main way to realize the customer segmentation is customer cluster. Design an efficient, accurate customer segmentation cluster algorithm, which is the key to realize customer segmentation of automobile collaborative industrial chain platform. Based on the problems of slow convergence rate and difficulty on initial cluster center selection of mixed data cluster algorithm-FKP, GA-FKP algorithm is raised in this thesis. This algorithm use GA to search the initial clustering center needed by FKP algorithm, optimize the chromosome with FKP algorithm and avoid premature convergence.GA_FKP cluster mining algorithm is an improved algorithm of mixed data cluster, and it has generality. FIMABoFM frequent itemset find algorithm find the frequent itmeset from a new perspective, and make up for the shortcomings of existing algorithms. Automobile industrial chain collaborative platform-based BI system architecture realizes the seamless integration and program resource sharing to the original platform, minimizes the workload of system development, and provide a reference method to information system BI integration of other small-and-medium-sized enterprises更多还原