首页 数据挖掘在客户关系管理中的—免费毕业设计论文

数据挖掘在客户关系管理中的—免费毕业设计论文

举报
开通vip

数据挖掘在客户关系管理中的—免费毕业设计论文数据挖掘在客户关系管理中的—免费毕业设计论文 学院理学学士论文 数据挖掘在客户关系管理中的应用 1 学院理学学士论文 前 言 前 言 随着Internet 时代的到来,电子商务、企业信息化得到飞速发展,全球市场竞争的加剧和客户需求多元化的要求,给传统的商业经济模式带来了巨大的冲击。传统企业基于4P (product 产品,place 渠道,price 价格,promotion 促销) 的竞争模式已逐渐被基于客户关系的经营理念所取代,未来市场的竞争已演变成为服务手段的竞争,建立和发展长期的客户关系是企业成...

数据挖掘在客户关系管理中的—免费毕业设计论文
数据挖掘在客户关系管理中的—免费毕业 设计 领导形象设计圆作业设计ao工艺污水处理厂设计附属工程施工组织设计清扫机器人结构设计 论文 学院理学学士论文 数据挖掘在客户关系管理中的应用 1 学院理学学士论文 前 言 前 言 随着Internet 时代的到来,电子商务、企业信息化得到飞速发展,全球市场竞争的加剧和客户需求多元化的要求,给传统的商业经济模式带来了巨大的冲击。传统企业基于4P (product 产品,place 渠道,price 价格,promotion 促销) 的竞争模式已逐渐被基于客户关系的经营理念所取代,未来市场的竞争已演变成为服务手段的竞争,建立和发展长期的客户关系是企业成功的关键,因此在客户关系管理中运用数据挖掘与其相关的技术已成为企业赢得核心竞争优势的重要手段。它是强调以“客户关系一对一理论”为基础,旨在改善企业与客户之间的新型管理机制,同时也是包括一个组织机构判断、选择、争取、发展和保持 [1]客户所要实施的全部商业过程。 面对当今社会信息化数据的迅速增长,“数据的爆长性增长与知识的相对贫乏”已成为现在人类最大的问题,人类分析数据和从中提取有用信息的能力已远远不能满足实际需要,所以我们要用数据挖掘技术来解决这些问题。数据挖掘就是从数据中发现趋势或模式的过程,其目的就是通过对大量数据的分析从中发现人们先前不知道的、但又非常有用的新的信息。数据挖掘的应用领域非常广泛,包括商务管理、生产控制、市场分析、工程设计和科学探索等,在CRM系统中也发挥着十分重要的作用。数据挖掘从大量的客户数据中挖掘出隐含的、先前未知的、对决策有极大价值的知识和规则,并能够根据已有的信息对未发生行为做出结果预测,为企业经营决策、市场策划提供依据。 因为在现实中存在了多种因素(如缩短的交易周期、增加的交易成本、新增的商品和物流方式、贪婪的竞争对手等)共同作用增加了客户关系的复杂性,所以成功企业必须对每一个因素做出及时反映。此外,大家都知道市场不等人,今天的客户到了明天也许就不再是你的,与客户的交互也不想过去那样简单,客户和潜在客户希望用他们习惯的方式进行交互,因此这就要求人们在做决策时注意以下规则: (1)提供恰当的服务。 (2)给合适的人。 (3)在正确的时间。 0 学院理学学士论文 前 言 (4)通过正确的渠道。 提供恰当的服务要求商家同客户进行多种交互,了解客户需求,投其所好,并把不相关的商品的数目降到最少。给合适的人意味着客户的需求各不相同。交互活动要在一个高度细化的市场上进行,以满足不同的需要。要在正确的时间是因为同客户的交互是一个连续的过程。最后,通过正确的渠道是因为商家可以通过各种途径(如邮寄、Email、电话购物等)与客户交互。对于不同的客户,商家必须保持选择最有效的方式,这样才有助于客户交互。 我们所要做的就是,使用数据挖掘工具运用在客户关系管理 (CRM) 中,利用数据挖掘技术来帮助商家解决同客户交互遇到的各种问题,通过对客户的深入了解,可以确切地估算出利润和投资回报率,可以避免一些不必要的损失,给商家带来了赢利。 1 学院理学学士论文 摘 要 摘 要 本论文主要讨论的是数据挖掘在客户关系管理系统中的应用,运用数据挖掘的相关技术(如粗糙集、关联规则、决策树、遗传算法等),结合销售行业的市场经营分析及CRM系统开发,建立基于数据挖掘的客户关系管理系统---客户服务的解决方案,并进行了部分实现。 数据挖掘技术在客户关系管理领域中的应用主要有: (1)通过数据挖掘判断客户的价值,即客户细分; (2)通过数据挖掘发掘潜在客户,从而实现交叉销售,提高现有客户的价值; (3)通过数据挖掘分析客户的流失,预防潜在的客户流失; (4)通过数据挖掘保留忠实客户,优化客户关系。 我的主要工作就是界面的设计、集成,数据库设计以及数据的预处理, 建立基于数据挖掘的客户关系管理系统。 在当今竞争激烈的市场环境中,CRM 是企业增加生存能力、扩大竞争优势不可或缺的支持。只有能够更好的利用客户信息、满足客户需求,一个企业才能够获得更大的利益。数据挖掘正是指导企业更好地理解客户以及满足客户需求的工具,为用户进行客户关系管理提供决策的参考依据。论文研究具有一定的理论意义和实际应用价值,为客户关系管理提供了一种研究思路和分析 方法 快递客服问题件处理详细方法山木方法pdf计算方法pdf华与华方法下载八字理论方法下载 。 关键词:客户关系管理(CRM);数据挖掘(Data Mining);界面设计;数据库设计; 数据处理 I 学院理学学士论文 ABSTRACT ABSTRACT This text mainly discusses the application of data mining in the system of customer relationship management, which applies the related techniques of data mining (such as rough set, association rule, decision tree, genetic algorithms, etc.), and combines the market management analysis and the CRM system’s development of sale business, and builds up the system of customer relationship management based on data mining—customer service, and makes part of functions come true. The techniques of data mining can be mainly applied in the field of customer relationship management: (1) Judge customers’ value through data mining, customer classification. (2)Find potential clients through data mining, and then carry out cross selling to improve the value of clients. (3)Analyze the drain of clients through data mining, to prevent the drain of potential clients. (4)Reserve the faithful clients through data mining, to optimize the relationship between clients. My task is mainly to design and integrate the interface of the software, to do data preprocessing, and to set up the system of customer relationship management based on data mining. In today’s environment of competitive market, CRM is such a necessary support for enterprises that it can increase the survival ability, and widen competition edge. Only by meeting the demands of clients and making good use of clients’ information can the enterprise gain more profit. And data mining is that tool which can guide companies to understand the clients and meet their demands, and finally, provide reference for the decision of customer relationship management. The discussion of this paper has certain theory meaning and actual application value, and also provides a kind of thinking and analyzing method. Keywords: data mining;customer relationship management;design of interface; database design;data processing II 学院理学学士论文 目 录 目 录 前 言.................................................. 0 摘 要.................................................. I ABSTRACT ................................................ II 第一章 可行性研究报告 .................................. 1 1.1项目开发的背景 .................................... 1 1.2系统设计相关的原理 ................................ 1 [2]1.2.1 VC ++6.0编程语言的介绍 ................... 1 [3]1.2.2 SQL Sever语句介绍 ........................ 2 [4]1.2.3 数据挖掘 (Data Mining) .................. 4 [5]1.2.4 客户关系管理 (CRM) ....................... 4 [6]1.2.5 数据挖掘在CRM中的应用 ................... 5 [7]1.3项目可行性评估 .................................. 6 1.3.1技术可行性评估 ............................... 6 1.3.2经济可行性评估 ............................... 6 1.3.3操作可行性评估 ............................... 6 1.4项目可行性研究报告 ................................ 6 第二章 系统需求分析 .................................... 7 2.1系统需要解决的主要问题 ............................ 7 2.1.1粗糙集 ....................................... 7 2.1.2关联规则 ..................................... 7 2.1.3决策树 ....................................... 8 2.1.4遗传算法 ..................................... 8 2.2系统应该具备的基本功能 ............................ 8 [4]2.3数据收集及预处理 ................................ 9 2.3.1对各算法中的数据进行预处理 .................. 10 [5]2.3.2数据仓库的定义和特性 ...................... 14 2.4数据字典 ......................................... 14 1 学院理学学士论文 目 录 第三章 系统设计 ........................................ 16 3.1前言 ............................................. 16 3.2基本简介 ......................................... 16 3.3系统功能模块结构图 ............................... 16 3.4运行环境 ......................................... 16 3.5详细分析 ......................................... 18 3.5.1 用户的权限限制 ............................. 18 3.5.2 数据库中表的设计 ........................... 18 3.5.3 功能模块的详细设计 ......................... 21 第四章 系统编码 ........................................ 23 4.1系统流程图设计 ................................... 23 4.2系统页面显示 ..................................... 24 4.2.1登入界面 .................................... 24 4.2.2进入客户资料管理界面 ........................ 26 4.2.3进入客户信息分析界面 ........................ 31 4.2.4用户信息管理界面 ............................ 33 第五章 系统测试 ........................................ 35 [7]5.1软件测试 ....................................... 35 5.2软件测试的目标与方法 ............................. 35 5.3模块测试 ......................................... 35 5.4集成测试 ......................................... 36 5.5 系统测试及实现 .................................. 36 第六章 开发难点与解决技巧 .............................. 37 6.1开发的难点 ....................................... 37 6.1.1各个算法模块间建立 .......................... 37 6.1.2界面与数据库的连接 .......................... 37 6.1.3两种不同用户的鉴别 .......................... 37 6.2系统的安全性 ..................................... 37 6.3系统的发布与维护 ................................. 38 参考文献 ................................................ 39 2 学院理学学士论文 目 录 致 谢 ................................................... 40 附 录 一 ................................................ 41 调研报告............................................. 41 Research Report ...................................... 43 附录二 .................................................. 46 操作手册............................................. 46 3 学院理学学士论文 第一章 可行性研究报告 第一章 可行性研究报告 1.1项目开发背景 本系统主要实现数据挖掘技术(如粗糙集、关联规则、决策树、遗传算法)在客户关系管理中的具体应用,其中粗糙集、关联规则算法由我们组的陈彩琴负责完成实现,决策树、遗传算法由吴海海负责完成实现,而我主要负责实现是界面的设计、数据库设计、数据预处理,把他们几个算法集成起来,建立一个基于数据挖掘的客户关系管理系统。 1.2系统设计相关原理 该客户关系管理系统使用VC++ 6.0作为前台开发工具,SQL Sever 2000 作为后台数据库,开发客户机/服务器(Client/ Sever ,缩写为 C/ S )的应用程序,此外还运用了Photoshop 7.0进行了图片处理。整体设计工程遵循软件工程的原理、技术和方法,经过需求分析、总体设计、文档和代码的编制、模块测试和系统实现几个阶段。下面就对这几中技术和方法做一个概述。 [2]1.2.1 VC ++6.0编程语言的介绍 Visual C ++ 是一个功能强大的可视化软件开发工具。它不仅是一个C++编译器,而且是一个基于Windows操作系统的可视化集成开发环境(integrated development environment, IDE)。Visual C ++ 6.0由许多组件组成,包括编辑器、编译器、调试器、以及程序向导AppWizard、类向导Class Wizard 等开发工具。这些组件通过一个名为Developer Studio 的组件集成为和谐的开发环境。 Visual C ++ 的特点: Visual C ++的源程序要求用C ++语言编写,它支持面向对象设计方法,并可以使用功能强大的软件基础类库MFC(Microsoft foundation class),充分体现了Microsoft公司的技术精华。用Visual C ++开发出来的软件稳定性好,可移植性强、而且软件与硬件相互独立。 利用Visual C ++可以编制各种类型的Windows 应用程序,从最简单的单文 1 学院理学学士论文 第一章 可行性研究报告 档和对话框程序到复杂的多文档和组合界面程序。并且Visual C ++作为Visual Studio 可视化组件家族中最重要的一个成员,它可以与其他可视化开发工具如Visual Basic、Visual J ++及Visual C#紧密地集成在一起,可进行不同类型的程序开发工作,适用于特殊、复杂和综合软件项目的开发以及系统软件的设计。 Visual C ++ 6.0源代码编辑器提供了自动语句完成的功能,编辑输入源程序时,它自动显示当前对象的成员变量和成员函数,并指明函数的参数类型。Visual C ++ 6.0的编译器增加了新的编译参数,改进了对ANSI C ++ 标准 excel标准偏差excel标准偏差函数exl标准差函数国标检验抽样标准表免费下载红头文件格式标准下载 的支持,并采用Microsoft的代码优化技术,使生成的目标代码更短小,应用程序运行的速度更快。Visual C ++ 6.0程序调试器功能更强大,它提供了诊断映射机制、无须要重编译的调试、远程调试和实现调试等功能。 Visual C ++ 拥有功能强大的MFC( Microsoft foundation Class Library)类库。MFC主要具有下面的优势: (1)MFC提供了大量的类,把Windows API函数封装在类中。用户使用MFC,可以很方便地开发Windows 应用程序,而不必为如何绘制Windows 窗口、对话框和菜单等常见用户界面发愁,因为它们都包含在MFC中; (2)因为可以充分利用前人开发的代码,所以缩短了程序的开发周期; (3)使代码具有更好的重要性; (4)使一些复杂的开发技术变得更容易实现,如网络编程和数据库编程等; (5)使用消息映射机制可以避免处理消息时的复杂情形; (6)使用MFC 可以提高运行效率。 迄今为止,Visual C ++ 可以说是一个功能最强大和使用范围最广泛的软件开发工具。 [3]1.2.2 SQL Sever语句介绍 SQL 全称是“结构化查询语言(Structured Query Language)”,最早的是IBM的圣约瑟研究实验室为其关系数据库管理系统SYSTEM R开发的一种查询语言,它的前身是SQUARE语言。SQL语言结构简洁,功能强大,简单易学,所以自从IBM公司1981年推出以来,SQL语言,得到了广泛的应用。如今无论是像Oracle ,Sybase, Informix, SQL server这些大型的数据库管理系统,还是像Visual FoxPro, PowerBuilder这些微机上常用的数据库开发系统,都支持SQL 2 学院理学学士论文 第一章 可行性研究报告 语言作为查询语言。 SQL的优点: 1.非过程化语言 SQL是一个非过程化的语言,因为它一次处理一个 记录 混凝土 养护记录下载土方回填监理旁站记录免费下载集备记录下载集备记录下载集备记录下载 ,对数据提供自动导航。SQL允许用户在高层的数据结构上工作,而不对单个记录进行操作,可操作记录集。所有SQL 语句接受集合作为输入,返回集合作为输出。SQL的集合特性允许一条SQL语句的结果作为另一条SQL语句的输入。 SQL不要求用户指定对数据的存放方法。 这种特性使用户更易集中精力于要得到的结果。所有SQL语句使用查询优化器,它是RDBMS的一部分,由它决定对指定数据存取的最快速度的手段。查询优化器知道存在什么索引,哪儿使用合适,而用户从不需要知道表是否有索引,表有什么类型的索引。 2.统一的语言 SQL可用于所有用户的DB活动模型,包括系统管理员、数据库管理员、 应用程序员、决策支持系统人员及许多其它类型的终端用户。SQL为许多任务提供了命令,包括: (1)查询数据 (2)在表中插入、修改和删除记录 (3)建立、修改和删除数据对象 (4)控制对数据和数据对象的存取 (5)保证数据库一致性和完整性 以前的数据库管理系统为上述各类操作提供单独的语言,而SQL将全部任务统一在一种语言中。 3.是所有关系数据库的公共语言 由于所有主要的关系数据库管理系统都支持SQL语言,用户可将使用SQL的技能从一个RDBMS转到另一个。所有用SQL编写的程序都是可以移植的。SQL是Postgre SQL(和大多数其它关系型数据库)用做查询语言的语言。它是可以移植的,并且容易学习使用。但是所有SQL语句都必须由数据库服务器独立地执行。这就意味着客户端应用必须把每条查询发送到数据库服务器,等待它处理这个查询,接收结果,做一些运算,然后给服务器发送另外一条查询。所有这些东西都会产生进程间通讯,并且如果客户端在另外一台机器上甚至还会导致网络开销。 3 学院理学学士论文 第一章 可行性研究报告 如果使用了 PL/pgSQL,那么可以把一块运算和一系列查询在数据库服务器里面组成一个块,这样就拥有了过程语言的力量并且简化SQL的使用,因而节约了大量的时间,因为用不着付出客户端/服务器通讯的过热。通过使用PL/pgSQL,应用可以获得可观的性能提升。 [4]1.2.3 数据挖掘 (Data Mining) 数据挖掘 (DM, Data Mining) 就是从大量的、不完全的、有噪声的、模糊的、随机的数据中,提取隐含在其中的、人们事先不知道的、但又是潜在的有用的信 息和知识的过程。 从商业角度来看,数据挖掘是一种新的信息处理技术,其主要特点是对商业数据库中的大量业务数据进行抽取、转化、分析和其他模型化处理,从中提取辅助商业决策的关键性数据。 要使数据挖掘在商业中起作用,就要把它与基本的商业活动过程关联起来。数据挖掘是企业同客户打交道的一长串步骤中的一个部分。数据挖掘影响商业活动的方式取决于商业过程,而不是数据挖掘的过程。 我们必须所指出的问题是: (1)数据挖掘的结果不同于其他数据驱动的商业活动。 (2)数据挖掘抽取的是数据库中不为用户所知的信息。 这就需要引进客户交互、情景。处理推销问题的用户要先理解数据挖掘的结果,然后才能把它运用到实际中。在实际应用中,如何使用数据挖掘的输出结果,最简单的办法是把结果看作黑盒。如果用户用黑盒给数据库中的客户评分,就能得到一张客户列表(可以向列表中的客户邮寄商品目录,提高他们的信用度)。用户所要做的只是把信封寄出去,这是非常有效的方法。 [5]1.2.4 客户关系管理 (CRM) 客户关系管理 ( Customer Relationship Management,简称 CRM )是选择和管理最有价值客户关系的一种商业策略。CRM 要求以客户为中心的商业哲学和企业文化来支持有效的营销、销售以及客户服务。如果企业拥有正确的领导、策 4 学院理学学士论文 第一章 可行性研究报告 略和企业文化,CRM 应用将为企业实现有效的客户关系管理。 客户关系管理首先是一种商业策略,它贯穿于企业的各个部门和经营环节,其目的在于理解、预测和管理企业现有的或潜在的顾客。客户关系管理涉及到战略、过程、组织和技术等各方面的变革,以使企业更好地围绕客户来有效地管理自己的经营。 我们为什么需要客户关系管理呢,因为大家都知道: (1)获得一个新客户比留住一个老客户要花更多的钱; (2)客户离开后再把他们吸引回来所花的钱要比他们一开始就满意所花的钱多得多; (3)将新的商品卖给老客户要比卖给新客户容易得多; (4)有些客户比其他客户更有利可图。有些客户现在是无利可图的,还有一些客户现在和将来都是无利可图的。 所以我们要运用新的技术,来改变单一的销售模式,寻求最有利可图的途径建立和维持稳定的客户关系。 客户对企业的好感和忠诚不仅来自于企业提供的商品,更来自于服务和经验等非实体因素。企业要了解客户的喜好---不仅包括产品,还包括样式和服务的外观。他们要管理每一个客户的关系,从每一个客户身上获取最大利润,降低市场营销费用,减少由于客户离去和无效的营销策略产生的浪费。以上都希望用客户关系管理的方法来实现。 [6]1.2.5 数据挖掘在CRM中的应用 在客户关系管理生命周期的各个阶段都有可能会用到数据挖掘技术。例如,在如何更好地为现有的客户提供服务方面,通过数据挖掘,能够帮助企业确定现有的客户的特点,从而为客户提供有针对性的服务;如何开拓企业新市场,获取新客户方面,通过数据挖掘,可以发现购买每一商品的客户的特征,从而可以向那些具有这些特征却没有购买的客户推销这个商品;如何挖掘现有客户群的潜力,为企业提供更多价值方面,通过数据挖掘,可以分析各个客户对企业带来的赢利能力,找到那些“金牌客户”后提供更加有针对性的服务,同时也可以分析现有客户的兴趣,向他们推荐提供新的产品和服务;在如何保持现有客户群方面,通过数据挖掘,可以找到流失客户的特征,以抢在那些具有相似特征的客户还未 5 学院理学学士论文 第一章 可行性研究报告 流失之前,就可以采取针对性的措施防止他们的流失;在预测公司未来发展方面,通过数据挖掘,可以预测客户赢利能力的变化,帮助企业提前采取各种对应的措施。 [7]1.3项目可行性评估 1.3.1技术可行性评估 根据现有的技术,采用Visual C ++、SQL Sever等技术和工具完全可以快速、高效地实现该系统模型。 1.3.2经济可行性评估 由于采用Visual C ++ 6.0 开发,用VC+ODBC连接数据库的开发方式尽可能地缩短开发周期、节约系统的开发费用,适合大、中型企业,适合远程网络,特别适合小型企业,操作简单、明朗,将会给它们带来丰厚的利润。 1.3.3操作可行性评估 可以参照其他的软件应用系统,特别适合管理级人员的使用,界面上能实现对客户资料的管理,为用户做出合理的决策,提供了很好的帮助,并尽可能地减少让用户难以操作的方式。 1.4项目可行性研究报告 纵上所述,从开发的语言、所要运用的工具,到数据挖掘在客户关系管理中应用的价值,都具备了可以开发本系统的要求。而且无论从技术、经济还是操作上,本系统的实现应该没有什么问题的,可以进行对系统的需求分析工作。 6 学院理学学士论文 第二章 系统的需求分析 第二章 系统需求分析 要实现一个软件系统,首先应该进行需求分析,这样才能设计出的软件满足用户的各项功能。下面就对客户管理系统的设计进行需求分析。 2.1系统需要解决的主要问题 首先,该系统主要是面向管理级人员而开发的,并且系统里的资料是企业客户的重要信息,所以用户者必须验证身份才能登入。用户根据不同的级别进入不同的模块,调用各自所需要的信息。本系统由二大模块组成——客户资料管理和客户信息分析,其中客户信息分析里引用了数据挖掘中的四个算法——粗糙集、关联规则、决策树、遗传算法。我所要实现的就是把这几个模块有效地结合起来,组成一个完整的模块,应用到客户关系管理系统中去。在这里,我来介绍以下的几种算法的需求分析。 2.1.1粗糙集 面对信息化数据的飞速发展,客户资料数据也从数以几十甚至上百 GB 的容量。这些数据是企业的财富,它如实地记录着企业运作的本质状况。但当客户数量达到一定数量时,企业已无能力对数量以外的人进行同样的销售努力,这就迫切需要一种分析高价值的客户的方法---粗糙集算法。它把数据集看作决策表,由用户指定数据库中某一个或多个属性作为分类的类别,并对决策表进行属性约简,根据这些属性的不同取值,将数据库中数据分为不同的类别,并对决策表进行属性约简和值约简,删除冗余的属性和属性值,在决策表中导出决策规则,从而从少量的客户属性资料数据判断谁是高价值客户,谁是低价值客户,在决策过程中应给予不同重视,那将给企业带来了无限商机。 2.1.2关联规则 公司与其客户之间的商业关系是一种持续的不断发展的关系。这就需要我们 7 学院理学学士论文 第二章 系统的需求分析 要用关联规则算法,来实现“交叉销售”。交叉销售分析就是从分析现有客户的购买行为数据开始,它是指同时向一位顾客销售多种相关的服务或产品,这是一种发现顾客多种需求,并满足其多种需求的一对一营销方式。交叉销售是建立在“双赢”原则的基础之上的,客户因得到更多、更好符合个性需求的服务而满足,企业也因销售的增长而获益。 2.1.3决策树 随着行业中竞争愈来愈激烈和获得一个新客户的开支愈来愈大,保持原有客户的工作也愈来愈有价值。比如说,在移动电话行业中,吸引并使一个新客户签约的开支估计为300美元或更多,这其中包括了硬件折旧和销售费用。然而保留一个旧客户的开支可能只是打一个电话或将他们的手机更新到使用最新的技术的那种型号。虽然这也很贵,但比起和一个全新的客户签约还是便宜了许多。如同在这个案例中看到的一样,我们看到可以用决策树来判断客户的流失,这在保留客户的活动中起着重要作用。 2.1.4遗传算法 自然界的生物体在遗传、变异和选择的相互作用下,优胜劣汰,不断地由低级向高级进化和发展。这种“物竞天择,适者生存”地进化规律实际上表示了一种优化的思想。在客户关系管理系统中,我们所实现的对客户优化的问题,也正是运用了这种思想。遗传算法用于CRM系统中模拟自然界的生物进化过程。它通过数据对客户群分析,运用适者生存的原则,对客户群进行分类,将没有效益的客户群或低效益客户群转化为高效益客户群,同时奖励和发展有效益的客户群。 2.2系统应该具备的基本功能 (1)用户权限的管理:用户可以根据不同的级别进行不同的管理,管理员可以看到整个系统,而部门管理员只能进行该部门的管理。 (2)客户基本资料管理:可以对客户进行增加、修改、删除、还可以进行多种查询功能。 8 学院理学学士论文 第二章 系统的需求分析 (3)判断客户属性资料:可以对客户属性资料进行管理,对粗糙集算法中的表的数据而进行增加或删除的功能。 (4)发掘潜在客户资料:可以对客户与商品之间的关系进行管理,对关联规则算法中的表的数据而进行增加或删除的功能。 (5)分析客户流失资料:可以对客户流失资料进行管理,对决策树算法中的表的数据而进行增加或删除的功能。 (6)判断客户价值:实现粗糙集算法的应用,对一个客户进行价值判断。 (7)发掘潜在客户:实现关联规则算法的应用,用关联规则推出哪些商品是关联的,然后导出哪些是潜在客户。 (8)客户流失分析:实现决策树算法的应用,用决策树来判断客户的流失,分析流失的原因。 (9)客户价值优化:实现遗传算法的应用,进行客户的优化。 除了实现上述功能以外,客户管理系统还应该具有友好、简洁的界面,安全性要高,稳定性要强。 [4]2.3数据收集及预处理 对于数据挖掘,我们当然需要收集大量的数据,收集了销售业和零售业的数据,还有网上的电线公司的数据 ,为了操作把这些数据集成于数据仓库中。 对于数据挖掘中的算法对其处理的数据集合一般都有一定的要求,比如数据完整性好、数据的冗余性少、属性之间的相关性小。然而,实际系统中的数据一般都具有不完全性、冗余性和模糊性,很少能直接满足数据挖掘算法的要求。另外,海量的实际数据中无意义的数据很多,严重影响了数据挖掘算法的执行效率,而且由于其中的噪声干扰还会造成无效的归纳,所以这就要求我们必须进行数据预处理这一步。 数据挖掘中的预处理主要是接受并理解用户的发现要求,确定发现任务,抽取与发现任务相关的知识源,根据背景知识中的约束性规则对数据进行检查,通过清理和归纳等操作,生成供挖掘核心算法使用的目标数据。 数据预处理应该包括以下几个方面的功能: 1.数据集成(Data Integration) 9 学院理学学士论文 第二章 系统的需求分析 数据集成主要是将多文件或多数据库运行环境中的异构数据进行合并处理,解决语义的模型性。该部分主要涉及数据的选择、数据的冲突问题以及不一致数据的处理问题。 2.数据清洗(Data Cleaning) 数据清洗要去除原数据集中的噪声数据和无关数据,处理遗漏数据和清洗脏数据,去除空白数据域和知识背景上的白噪声,考虑时间顺序和数据变化等。主要包括重复数据处理和缺省值数据处理,并完成一些数据类型的转换。 3.数据变换(Data Transformation) 数据变换主要是找到数据的特征表示,用维变换或转换方式减少有效变量的数目或找到数据的不变式,包括规格化、归纳、切换、旋转和投影等操作。 4.数据简化(Data Reduction) 数据简化是在对发现任务和数据本身内容理解的基础上,寻找依赖于发现目标的表达数据的有用特征,以缩减数据规模,从而在尽可能保持数据原貌的前提下最大限度地精简数据量。它主要有两个途径:属性选择和数据抽样,分别针对数据库中的属性和记录。 (1)属性选择包括针对属性进行剪枝、并枝、找方程和找相关属性等操作。 (2)数据抽样就是进行数据记录之间的相关性分析,用少量的记录的线性组合来表示大量的记录。 2.3.1对各算法中的数据进行预处理 (1)对粗糙集数据的进行预处理。 首先,Excel表的数据转换到SOL Sever 中要考虑到数据类型的选择问题,应尽量选择占物理空间较小的数据类型,如在区域范围内用varchar 替代CString 类型,因为varchar可以灵活的分配所要的空间,对于大规模数据集来说将会大大减少系统开销。 对年龄、收入属性的不同取值(区间)数小于对应的属性概化阀值,则应对它们进行概化。如年龄,它的概化阀值设为20,所以小于30岁的,我们设为young,30~50的为middle,大于50的为senior。同理,把收入小于2000的设为low,大于、等于2000为high。 由于婚否属性只有两个不同值,所以该属性保留,并且不对其进行概化。同 10 学院理学学士论文 第二章 系统的需求分析 理,性别、是否全职与单位属性也被保留,也不需要进行概化。 数据处理前:如2-1所示。 表2-1 原始数据 Chart 2-1 the Origin Data 数据在SQL Sever中进行处理以后:如表2-2所示 表2-2 处理后的数据 Chart 2-2 the Final Data 11 学院理学学士论文 第二章 系统的需求分析 (2)对关联规则数据的进行预处理。 同上,我们在SQL Sever里用的数据类型还是varchar,但为了方便对关联规则算法实行,还要对所需要的商品进行符号替换。用I1代表随身听,同理:I2耳机,I3电池,I4笔筒,I5充电器,I6 笔记 哲学笔记pdf明清笔记pdf政法笔记下载课堂笔记下载生物化学笔记PDF 本,I7钢笔,I8手表,I9MP3,I10耳麦,I11CD机,I12音响,I13电脑,I14手机,I15键盘,I16 鼠标,I17摄像头,I18CD包,I19钱包,I20背包等。 数据处理前:如表2-3所示 表2-3 原始数据 Chart 2-3 the Origin Data 数据在SQL Sever中进行处理以后:如表2-4所示 表2-4 处理后的数据 Chart 2-4 the Final Data 12 学院理学学士论文 第二章 系统的需求分析 (3)对决策树数据的进行预处理。 同上,我们在SQL Sever里用的数据类型还是varchar。对于入网时间、服务时间存在大量不同值,应当进行概化。如入网时间,假定存在概念分层,为了方便算法的运行,可以将它从(年/月/日)概化到年,而年又可以概化到年段,并且年段的不同值(区间)数小于对应的属性概化阀值,则应当对它进行概化。假设入网时间的概化阀值为1,则它的分段为96,97,98,99。同理,服务时间分为三段:小于20, 20 ~50之间,大于50。 对年龄、掉线次数属性的只需进行概化。假设年龄概化阀值设为10,则它的分段为: 小于30,30~40之间,大于40。同理,掉线次数分为:0,1~2之间,大于2。 由于话费类型属性只有两个不同值,所以该属性保留,并且不对其进行概化。同理,性别、手机品牌属性也被保留,也不需要进行概化。 数据处理前如表2-5所示,数据在SQL Sever中进行处理以后如表2-6所示。 表2-5 原始数据 Chart 2-5 the Origin Data 表2-6 处理后的数据 Chart 2-6 the Final Data 13 学院理学学士论文 第二章 系统的需求分析 [5]2.3.2数据仓库的定义和特性 正式的数据仓库定义是W.H.Inmon提出的,它是一个面向主题的、集成的、不同时间的、稳定的数据的集合,它支持管理决策。换句话说,数据仓库包括: (1)一种或多种从任何类型的数据结构(包括平面的、层次的、关系的或对象等类型;包括开发的或专用的,内部的和外部的)中抽取字段的工具; (2)利用数据源“目录”把数据综合到一个稳定的、集成的、面向主题的数据库中; 数据仓库应具有的特性: (1)一个用于分析任务的数据库,分析的数据来自多个应用程序; (2)支持少数几个用户长时间的交互操作; (3)它的内容周期性更新(大部分情况是增加数据); (4)它包含当前的和过去的数据,能对信息做历史性的透视分析; (5)包含几个大的数据表; (6)每个查询会频繁产生大的结果集,还包括对整个表的频繁遍历和多个表的 联结操作。 2.4数据字典 数据字典是系统分析阶段的重要文档,它清楚的定义与详细的解释了数据流图上未能详细表达的内容,同时它能够帮助分析员组织数据,是用户交谈的好工具。此外,数据字典记载着每个数据项的关键条目,从中可以检查出各数据元素的详细内容或漏掉的数据元素。本系统中的数据字典主要有: 用户表:记录不同的用户,与相对应的部门和密码; 客户资料表:记录不同的客户,与相对应的客户个人信息,方便与客户进行沟通、联系; 粗糙集表:记录客户的有关属性,从这些属性来进行属性约简和值约简的判断; 关联规则表:记录着客户消费的有关数据记录,买了哪些商品,商品间存在着哪些强关联规则; 决策树表:记录着客户消费的类型,分析客户的流失; 14 学院理学学士论文 第二章 系统的需求分析 粗糙集数据测试表:用来检测粗糙集算法的正确性; 关联规则数据测试表:用来检测关联规则算法的正确性; 决策树数据测试表:用来检测决策树算法的正确性。 15 学院理学学士论文 第三章 系统设计 第三章 系统设计 3.1前言 经过需求分析阶段的工作,系统必须“做什么”已经清楚了,现在是决定“怎样做”的时候。总体设计的基本目的就是回答“从总体上说,系统应该如何实现,”这个问题。因此,总体设计又称为概要设计或初步设计。通过这个阶段的工作将划分出组成系统的物理元素---程序、文件、数据库、人工过程和文档等等,但是每个物理元素仍然处于黑盒子级,这些黑盒子里的具体内容将在以后进行详细的设计。总体设计阶段的另一项重要任务是设计软件的结构,也就是要确定系统中每个程序是由哪些模块组成的,以及这些模块相互间的关系。 3.2基本简介 该系统是主要针对于管理类人员使用,分为客户资料管理模块、客户信息分析两大模块。它应该具有稳定性、保密性、方便性和灵活性。系统管理员可以轻松地进行增加、修改、删除和查询客户资料,以及对每个算法所用的表,可以进行增加或删除数据的功能,对客户信息分析的模块也有较好的了解。而部门管理员可以了解到本部门所要的信息,方便查阅、分析信息,为更好地做好销售而服务。 3.3系统功能模块结构图 根据系统功能的要求,客户关系管理系统主要分为二大模块,即客户资料管理模块、客户信息分析模块。可以将本系统分解成几个功能模块分别设计,功能模块之间的关系图,如图3-1所示。 3.4运行环境 为了保证系统运行的效率和可靠性,系统服务器端与客户端的要求不能很低。该系统可单机允许,也可适用于内部的局域网。其运行要求如下: 16 学院理学学士论文 第三章 系统设计 客 户 管 理 管 理 系 统 系 统 管 理 客户资料管理 客户信息分析 判发客客关系管理系统 权用修客判发查分断掘户户限 户改户断掘询析 客潜价流限信密基客潜客户在值失 制息码本户在户 价客优分管资价客流值户化 析 理料值户失 资资资 料料料 粗关决遗 糙联策传 集规数算 算则算法 法算法 法 图3-1系统功能模块图 Fig 3-1 Systematic Function Module 软件环境: 操作系统:Microsoft Windows 98及以上版本 编程环境:Visual C++ 6.0 数据库:采用SQL Sever,运行于服务器端 17 学院理学学士论文 第三章 系统设计 硬件环境: 服务器 CPU:PIII 500以上 ,内存:128M以上 客户机 CPU:P200MMX以上,内存:32M以上 3.5详细分析 详细设计阶段的根本目标是确定应该怎么具体地实现所要求的系统,是整个设计过程中最重要的步骤之一。下面就分如下几个部分对系统进行详细设计:(1)用户的权限限制 (2)数据库中表的设计 (3)功能模块详细设计。 3.5.1 用户权限限制 管理员:在此系统中可以由若干个管理员,即系统管理员。他们具有一般用户没有的权限,即拥有整个系统的管理,可以看到整个系统的信息,还可以对部门管理员进行添加或删除。 一般用户:是相对于部门来讲,即部门管理员。他们只能看到自己部门的信息,分析所要的信息。 3.5.2 数据库中表的设计 1.用户登入表结构 表3 -1用户登入表 Chart 3-1 User Login 各自段说明: (1) UserName 字段表示用户所输入的名称,设置为主键,它代表是唯一 的。 (2) Pwd 字段表示每个用户所对应的密码。 18 学院理学学士论文 第三章 系统设计 (3) User_type 字段表示每个用户所属的类型。 2.客户信息表结构 表3-2客户信息表 Chart 3-2 Customer Information 各字段的说明: (1)Customer_id 字段表示客户的ID号,设置为主键,它代表是唯一的。 (2)name 字段表示客户的名字。 (3)address 字段表示客户的地址。 (4)code 字段表示客户的邮编。 (5)profession 字段表示客户的职位。 (6)email 字段表示客户的电子邮件地址。 (7)phone 字段表示客户的电话号码。 (8)mobile 字段表示客户的手机号码。 (9)meet_time 字段表示 会见客户的时间。 (10)contract_name 字段表示接待客户的联系人。 (11)memo字段表示客户的备注。 (12)company 字段表示客户所属哪个公司。 3(关联规则表结构 表3-3 关联规则表 Chart 3-3 Association Rule 19 学院理学学士论文 第三章 系统设计 各字段的说明: (1)TID 字段表示客户的 ID号,代表每一个客户。 (2)Item 字段表示每一个客户所买的商品。 4(决策树表结构 表3-4 决策树表 Chart 3-4 Decision Tree 各字段的说明: (1)Customer_id 字段表示客户的ID号。 (2)Sex 字段表示客户的性别。 (3)Age 字段表示客户的年龄。 (4)In_time 字段表示客户的入网时间。 (5)Service_time 字段表示客户的服务时间。 (6)Drop_number 字段表示客户的掉线次数。 (7)Pay_way 字段表示客户的支付方式。 (8)Cost_type 字段表示客户的话费类型。 (9)Phone_type 字段表示客户的手机品牌。 (10)Whether_lost 字段表示客户的是否流失。 5(粗糙集表的结构 表3-5粗糙集表 Chart 3-5 Rough Set 20 学院理学学士论文 第三章 系统设计 各字段的说明: (1)U 字段表示客户的ID 号。 (2)a1 字段表示客户的年龄。 (3)a2 字段表示客户的受教育程度。 (4)a3 字段表示客户的收入情况。 (5)a4 字段表示客户的婚姻状况。 (6)a5 字段表示客户的性别。 (7)a6 字段表示客户的是否全职。 (8)a7 字段表示客户的单位属性。 (9)d 字段表示客户的价值。 3.5.3 功能模块的详细设计 下面对各个功能模块分别进行详细讨论,具体情况详见系统源程序代码。 3.4.3.1用户登入模块 用户根据自己的密码登入系统,系统根据不同的用户来判别它属于哪个部门,进而判别它拥有哪些权限。 3.4.3.2用户信息管理模块 管理员可以添加各个不同部门的用户,同时也可以删除不再需要的用户。同时,管理员也可以浏览整个系统中的各种用户。各个用户也可以修改自己的密码。 3.4.3.3客户资料管理模块 只有管理员才可以进入该模块,可以根据需要对客户资料进行增加、修改、删除,还可以按照不同的需要对客户资料进行查询,方便与客户沟通、联系。管理员还可以对客户信息分析的模块的数据进行增加或删除的权利。 3.4.3.4客户信息分析模块 客户信息分析模块是整个系统的核心模块,所有的用户都可以进入该模块。它主要实现用数据挖掘的四个算法(粗糙集、关联规则、决策树、遗传算法)在客户关系管理系统中的应用。粗糙集算法实现对客户价值的判断,为客户评分;关联规则算法实现发掘潜在客户,使用交叉销售提高现有客户的价值;决策树算法 21 学院理学学士论文 第三章 系统设计 实现客户流失分析,为保留客户活动而努力;遗传算法实现对客户价值优化,保留忠实客户。 22 学院理学学士论文 第四章 系统编码 第四章 系统编码 在进行了系统设计以后,就开始进行编码工作。也就是设计文件和把他们有机地组织在一起实现所有功能。 4.1系统流程图设计 整个系统的流程分析图如图4-1所示。 开 始 重 试 否 是 失 败 用户登入 退出程序 失败超过3次 成 功 判断客户价值资料 读取用户类型 发掘潜在客户资料 分析客户流失资料 判断客户的价 管理部用户 客户基本信息 值 发掘潜在客户 1) 管理用户信息 2) 修改自己的密码 分析客户流失 客户价值优化 销售部用户 修改自己的密码 帮助文件 图4-1系统流程分析图 Fig 4-1 Systematic flow analysis 23 学院理学学士论文 第四章 系统编码 4.2系统页面显示 4.2.1登入界面 用户登入界面,如图4-2所示。 图4-2 用户登入界面 Fig 4-2 the Page of User Login 这个登入界面实现很简单,但它要调用的SQL Sever 数据库中的数据表:如表4-1所示。 表4-1 用户登入表 Chart 4-1 User Login 登入界面实现的主要函数代码如下: bool CKHApp::IsEnter() { extern CString alluser; extern CString alltype; 24 学院理学学士论文 第四章 系统编码 CKHSet *m_pSet = new CKHSet; m_pSet->Open(); int count = 0; start: if(count >=3) { AfxMessageBox("输入已有三次之多,强迫退出!"); return false; } EnterDlg dlg; m_pSet->MoveFirst(); if(dlg.DoModal() == IDOK) { if(dlg.m_PassWord == ""||dlg.m_UserName == "") { goto start; } CString user,password,type; user = dlg.m_UserName; password = dlg.m_PassWord; alluser = user; while(!m_pSet->IsEOF()) { if(user==(CString)m_pSet->m_UserName&&password ==(CString) m_pSet->m_Pwd) { alltype = (CString)m_pSet->m_User_type; alluser = user; m_pSet->Close(); 25 学院理学学士论文 第四章 系统编码 return true; } else { m_pSet->MoveNext(); continue; } } AfxMessageBox("用户和密码可能错误!"); count ++; goto start; m_pSet->Close(); } else { AfxMessageBox("你要退出吗!"); return false; } } 全部代码的实现是在VC中,它实现了对进入用户的判别,分别属于哪个部 门,进行着哪些权限。用户登入的密码错误不能超过3次,如果超过,系统将会 自动关闭。 4.2.2进入客户资料管理界面 (1)客户基本资料界面 客户基本资料界面,如图4-3所示。 26 学院理学学士论文 第四章 系统编码 图4-3 客户基本信息界面 Fig 4-3 the Page of the Basic Information of Client 这个界面的实现同样调用在SQL Sever 数据库中的数据表:如表4-2所示。 表4-2 客户信息表 Chart 4-2 Client Information 客户信息管理界面实现的主要代码如下(以系统中源代码为主): BOOL CKehuDlg::OnInitDialog() { 27 学院理学学士论文 第四章 系统编码 CDialog::OnInitDialog(); // TODO: Add extra initialization here InitControl(); m_pSet->Open(); while(!m_pSet->IsEOF()) { int nIndex = m_listCustomer.GetItemCount(); CString getid = (LPCTSTR)m_pSet->m_Customer_id; getid.Format("%d",m_pSet->m_Customer_id); LV_ITEM lvItem; lvItem.mask = LVIF_TEXT; lvItem.iItem = nIndex; lvItem.iSubItem = 0; lvItem.pszText = (char *)(LPCTSTR)getid; m_listCustomer.InsertItem(&lvItem); m_listCustomer.SetItemText(nIndex,1,m_pSet->m_name); m_listCustomer.SetItemText(nIndex,2,m_pSet->m_address); m_listCustomer.SetItemText(nIndex,3,m_pSet->m_code); m_listCustomer.SetItemText(nIndex,4,m_pSet->m_profession); m_listCustomer.SetItemText(nIndex,11,m_pSet->m_company); m_listCustomer.SetItemText(nIndex,5,m_pSet->m_email); m_listCustomer.SetItemText(nIndex,6,m_pSet->m_phone); m_listCustomer.SetItemText(nIndex,7,m_pSet->m_mobile); m_listCustomer.SetItemText(nIndex,8,m_pSet->m_meet_time); m_listCustomer.SetItemText(nIndex,9,m_pSet->m_contract_name); m_listCustomer.SetItemText(nIndex,10,m_pSet->m_memo); m_pSet->MoveNext(); } 28 学院理学学士论文 第四章 系统编码 m_pSet->Close(); return TRUE; // return TRUE unless you set the focus to a control // EXCEPTION: OCX Property Pages should return FALSE } 只有管理部的人员才可以进入客户基本信息界面,对客户信息进行管理,如增加、删除或修改哪些客户,可以根据不同的要求对客户信息进行查询。要保持与客户良好的沟通,及时反应到信息之中,为以后的销售做更好的服务。 (2)判断客户价值资料界面 判断客户价值资料界面,如图4-4所示。 图4-4判断客户价值资料界面 Fig4-4 the Page of the Datum of Judge Client value 界面的实现同样要调用在SQL Sever 数据库中的粗糙集表,见表2-2。 这个界面同样也只有管理部的人员才可以进入此界面,它主要是对粗糙集算法所用的表进行管理,可以增加所要的属性或删除多余、重复的属性,来保证粗糙集算法可以更好地进行属性约简和值约简。 (3)发掘潜在客户资料界面 发掘潜在客户资料界面,如图4-5所示。 29 学院理学学士论文 第四章 系统编码 图4-5发掘潜在客户资料界面 Fig4-5 the Page of the Datum of Dig the Future Client 界面的实现同样要调用在SQL Sever 数据库中的关联规则表:见表2-4。 这个界面同样只有管理部的人员才可以进入此界面,它主要是对关联规则算法所用的表进行管理,可以对表里的数据进行增加或删除操作。 (4)分析客户流失资料界面 分析客户流失资料界面,如图4-6所示 图4-6分析客户流失资料界面 Fig4-6 the Page of the Datum of Assay the Lapsing Client 界面的实现同样要调用在SQL Sever 数据库中的关联规则表:见表2-6。 30 学院理学学士论文 第四章 系统编码 这个界面同样只有管理部的人员才可以进入此界面,它主要是对决策树算法所用的表进行管理,可以对表里的数据进行增加或删除操作。 4.2.3进入客户信息分析界面 (1)判断客户价值界面 判断客户价值界面,如图4-7所示。 图4-7判断客户价值界面 Fig4-7 the Page of Judge Client value 这个界面的实现同样调用在SQL Sever 数据库中的粗糙集表,见表2-2。 判断客户价值界面的实现主要运用粗糙集算法的实现。粗糙集算法根据客户的不同属性来取值,将数据库中数据分为不同的类别,并对表进行属性约简和值约简,删除冗余的属性和属性值,在表中导出决策规则,从而就可以从少量的客户属性资料数据就可以判断是高价值客户或低价值客户,为客户评分,可以进行促销活动,提供了更多的商机。 (2)发掘潜在客户界面 31 学院理学学士论文 第四章 系统编码 发掘潜在客户界面,如图4-8所示。 图4-8 发掘潜在客户界面 Fig4-8 the Page of Dig the Future Client 界面的实现同样要调用在SQL Sever 数据库中的关联规则表: 见表2-4。 发掘潜在客户界面的实现主要是运用关联规则算法来实现。首先,要进行参数设置,对商品总数、支持度、置信度进行设置值。然后,根据这些数据来推出哪些组合商品是哪些客户要买的,由买下的商品导出可能被卖出的商品,在从这些商品中导出潜在客户,根据客户的ID号来查询客户的信息,与客户更好地联系,促进销售。 (3)客户流失分析界面 客户流失分析界面,如图4-9所示。 图4-9客户流失分析界面 Fig4-9 the Page of Assay the Lapsing Client 32 学院理学学士论文 第四章 系统编码 界面的实现同样要调用在SQL Sever 数据库中的决策树表,见表2-6。 客户流失分析界面的实现主要运用决策数算法来实现。决策数算法根据客户消费的属性来判断一个客户的流失,分析流失的原因,为尽可以的保留客户而努力,因为大家都知道要发展一个新客户所花的费用比留住一个老客户所花的费用要多的多,而且老客户更容易为企业带来利润,所以我们要尽量保持每一个客户。 (4)客户价值优化界面 客户价值优化价值,如图4-10所示。 图4-10 客户价值优化界面 Fig 4-10 the Page of privilege the Value Client 客户价值优化界面的实现主要运用遗传算法的实现。遗传算法要制定一个目标函数,根据它的交叉率和变异率进行优化,迭代计算,算出最优个体。根据它的最佳个体得出,客户关怀频率值与商业交往频率值,从而得出最优化的客户。 4.2.4用户信息管理界面 用户信息管理界面,如图4-11所示。 图4-11 用户信息界面 Fig 4-11 the Page of User Information 33 学院理学学士论文 第四章 系统编码 只有管理部的人员才可以进入用户信息管理界面,对用户信息进行管理,如增加新用户或删除不再需要的用户,还可以浏览每个用户的信息,属于哪个部门的,但不包括用户的密码。 34 学院理学学士论文 第五章 系统测试 第五章 系统测试 [7]5.1软件测试 软件测试就是在软件投入运行前,对软件需求分析、设计规格说明书和编码的最终复审,是软件质量保证的关键步骤。确切的说,软件测试就是为了发现错误而执行的过程。一般分为两个阶段。单元测试是在编完某一模块后进行测试;)综合测试是在开发完软件后进行综合测试。 5.2软件测试的目标与方法 G.Myers给出了关于测试的一些规则,这些规则可以看作是测试的目标或定义: (1)测试是为了发现程序中的错误而执行程序的过程; (2)好的测试方案是极可能发现迄今为止尚未发现的错误的测试方案; (3)成功的测试是发现了至今为止尚未发现的错误的测试。 测试任何软件都有两种方法:黑盒测试法和白盒测试法。 黑盒测试法(又称功能测试法)是把程序看成一个黑盒子,完全不考虑程序的内部结构和处理过程,是在程序接口进行的测试,它只检查程序功能是否能按照规格说明书的规定正常使用。 白盒测试法(又称结构测试法)是把程序看成装在一个透明的白盒子里,也就是完全了解程序的结构和处理过程,这种方法按照程序内部的逻辑测试程序,检验程序中的每条通路是否都能按照预定要求正确工作。通过测试并不能证明程序就是完全正确的,但是我们的目的就是要通过测试保证软件的可靠性,因此,必须仔细设计测试方案,力争用最少的测试发现尽可能多的错误。 5.3模块测试 模块测试也叫单元测试,它的依据是详细设计描述,模块测试应对模块内所有重要的控制路径设计测试用例。 测试环境为Winxp与Win2000。 35 学院理学学士论文 第五章 系统测试 5.4集成测试 由于各模块有较强的独立性,所以集成测试比较简单,重点测试了各模块共用的几个变量和与主窗体的接口,并对整体功能进行了测试。 5.5系统测试及实现 本系统先进行各模块测试,经过修改和调整通过以后,进行总体测试。测试结果各项功能均已经或基本达到设计要求。各主要界面在前面已经给出,可能还会有一些错误,这只有在以后的运行中加以改进。 36 学院理学学士论文 第六章 开发难点与解决技巧 第六章 开发难点与解决技巧 6.1开发的难点 6.1.1各个算法模块间建立 每个算法都是相应的是一个独立的模块,通过界面把它们有机地连接在一起,组成一个完整的整体。因为它涉及到与数据库的连接,所以每一部分都需要用到数据库的打开与断开语句,使得每一个模块都可以相应的独立运行。 6.1.2界面与数据库的连接 界面与数据库的连接,我们运用ODBC的连接方式。ODBC是一个应用广泛的数据库访问编程接口,使用标准的SQL作为其数据库访问语言。ODBC的设计是建立在客户机/服务器体系结构基础之上。它避免了与数据库连接的复杂性。我们只需要安装SQL Sever的ODBC的驱动程序,就可以建立应用程序和数据库系统的通信接口,从而是应用程序快速访问和操作数据库。 6.1.3两种不同用户的鉴别 采用两层用户鉴别的模式,目的是为了系统的安全,当然,这个办法所起到的作用也非常有限,只是能够防止普通用户通过猜测管理员的用户名和密码达到入侵的目的。 6.2系统的安全性 客户关系管理系统一般运用在局域网内,也可以在远程互联网上运用。使用身份验证机制保护来系统里的内容,系统可以要求每个试图访问者必须使用有效的Windows NT账号的用户名和密码。每当用户试图访问时,Web服务器将进行身份验证,即确认用户身份,以检查用户是否拥有有效的Windows NT账号,再进行密码确认。 37 学院理学学士论文 第六章 开发难点与解决技巧 6.3系统的发布与维护 本系统所进行的操作系统是在Windows 2000 server或Windows XP。其Web服务器使用Microsoft的Internet Information Server5.0,在发布本系统时首先确认服务器计算机安装了IIS5.0服务器软件。验证方法设置为Windows集成验证,集成Windows验证使用与用户的Internet Explore Web浏览器进行密码交换以确认用户的身份。要用SQL Sever作为后台的数据库。 38 学院理学学士论文 参考文献 参考文献 [1]王扶兵,李兵. CRM在客户关系分析评价方法研究[J]. 计算机工程与应用, 2003年第31期 [2]王育监. Visual C ++面向对象编程教程[M].北京:清华大学出版社,2003 [3]启明工作室.Visual C ++ +SQL Sever 数据库应用系统开发与实例[M].北京:人民邮电出版社,2004 [4]邵峰晶,于忠清.数据挖掘原理与算法[M].北京:中国水利水电出版社,2003 [5]Alex Berson.构建面向CRM 的数据挖掘应用[M].北京:人民邮电出版社,2001 [6]董金祥,陈钢,伊建伟.客户关系管理CRM[M].杭州:浙江大学出版社,2002 [7]张海藩.软件工程导论[M].北京:清华大学出版社,2003 39 学院理学学士论文 致谢 致 谢 经过我们组成员的共同努力,我们圆满地完成“数据挖掘在客户关系管理上的应用系统”,首先要感谢柳炳祥老师对我们的细心指导和督促,同时感谢系各位领导和各位老师给我们的关怀和督促,以及系资料室、系专业机房对我们的大力支持和帮助,为我们提供了一个良好的毕业设计环境。 要感谢同组成员陈彩琴、吴海海,还有我们班的李海林同学等人对我的帮助和配合,以及身边的同学给我的鼓励和帮助,让我得以顺利完成我的毕业设计任务。这一段的毕业设计对我以后的工作、学习和生活将有巨大的帮助和影响,是一笔宝贵的财富。但由于时间有限和技术上的一些难题,毕业设计还有很多不完善的地方,是今后需要进一步研究的工作。 40 学院理学学士论文 附录一 附 录 一 调研报告 数据挖掘在CRM中运用 (1)通过数据挖掘获得新的客户。 在CRM中首先应识别潜在客户,然后将他们转化为客户。Big Bank and Credit Card(BB&CC)公司每年通过邮递的方式开展25 次促销活动,每次给一百万人提供申请信用卡的机会,BB&CC 公司会将信用高的申请者接受为服务对象,最终只有1%的申请者成为用户。 BB&CC公司所面临的挑战是如何让邮递促销活动更加有效。首先,BB&CC公司抽取了一个50,000人的样本,做了一个测试。在样本测试结果分析的基础上建立了两个模型,一个用来预测谁将填写申请表(使用决策树方法),另一个是信用评估模型(使用神经网络方法)。从剩下的950,000 个人中再次抽取700,000个样本,使用模型找出哪些人会对促销活动做出反应,并且具有良好的信用。结果如下:包括建模型时用的50,000 共抽取了750,000个样本,其中9,000 个申请者被接受,接受率从1%上升到了1.2%。 数据挖掘虽然不能准确的识别哪10,000个申请者最终会成为用户,但是可以促使营销活动更加有效。 (2)通过数据挖掘使用交叉销售提高现有客户的价值。 Guns and Rouses(G&R)公司销售的产品是:仿迫击炮与大炮的室外花盆和仿大口径手枪与长枪的室内花盆。产品表被发往12,000,000个家庭。当客户电话定购某个产品时,(G&R)公司会积极的推销其它的产品——交叉销售。但是,(G&R)公司发现只有1/3的客户允许他们提出建议,最终的交叉销售率不足1%,并招致了一片抱怨声。为此B&R公司想确定到底是哪些人在定购某个产品的同时需要其他的产品。 G&R公司建立了两个数据挖掘模型,一个是用来预测某个客户是否会被建议触怒,另一个用来预测什么样的建议会被很好的接受。数据挖掘模型使用客户 41 学院理学学士论文 附录一 信息数据库中客户的信息和新的客户信息,告诉销售代表哪种人可以采用交叉销售的方式以及建议什么产品。交叉销售的成功率上升到了2%,而且很少有抱怨。 数据挖掘帮助B&R公司更好的了解了客户的需求。当把数据挖掘模型与典型的CRM交叉销售活动结合起来时,B&R 公司的景况完全改变了。 (3)通过数据挖掘提高现有客户的价值 Big Sam’s Clothing 是一家制衣公司,它建立了一个网站,提供了友好的用户登陆界面;进行在线销售。一旦你定购了某件商品或注册为成员后,公司将会通过电子邮件的方式向你推荐一些你可能感兴趣的新产品F当你在网站上查阅一件产品时,网站会向你推荐一些与该产品有关的你可能感兴趣的产品。 当公司最初建立这个网站时,根本没有考虑个性化,仅仅将它作为一个产品目录的在线版。借助数据挖掘工具公司改善了网站的结构,使用聚类方法将以往按产品类型分类替换成按互补原则分类。当客户寻找某个商品时便可以使用这些分组提供建议,然后建立客户数据,识别哪些客户可能对产品目录中添加的新产品感兴趣。同时,Big Sam’s 公司用数据挖掘预测那些会引起客户兴趣的新产品,并通过电子邮件传递给客户,客户可以选择是否接收。 Big Sam’s 公司确定了个性化战略,通过提供主动性客户服务指导客户挑选商品,不仅增加了销售额而且加强了与客户之间的联系。客户已经将它看作购物时忠实的顾问。 (4)通过数据挖掘保留忠实客户。 Know Service公司是一家网络服务公司,像其它的公司一样也在不停地流失客户。流失率每月8%,这意味着现有的一百万个客户中每个月会有80,000个客户流失。Know Service公司重新寻找一个客户的成本是$200,每个月公司要在寻找新客户上投资$16,000,000,因此Know Service公司需要数据挖掘的帮助。 Know Service公司掌握了客户的大量上网信息和客户的个人信息,构建了客户信息数据库。Know Service公司做的第一件事就是从客户数据库中选择、转变数据。接下来所作的就是根据客户支出、生命周期对客户进行划分,并判定客户的持久性,识别哪些客户在将来会成为忠实的客户然后在忠实的客户中识别哪些有可能流失。第三步,确定最优服务、保持客户忠实。例如对上网支出大的客户可以提供包月的服务而不是提供更加大的主页空间。 结果客户流失率从原来的8%下降到7.5%,每个月节省了$1,000,000。 42 学院理学学士论文 附录一 Research Report The data mining is using in the CRM (1)Acquire the new customers through data mining. Should identify the latent customer first in the CRM, and then convert them as the customer. Annually the Big Bank and Credit Card (BB& CC) company opens the exhibition in the way of postal delivery to promote sales the activity 25 times, the every time gives 1,000,000 people the opportunity that provides to apply for the credit card, the BB& CC company meeting accept high applicant of reputation for the service object, end only have 1% applicant to become the customer. How the challenge that the BB& CC company face is let the postal delivery promotes sales the activity more valid. First, the BB& CC Company sampled the sample of a 50,000 people, doing a test. Test the analytical foundation of result in the sample up built up two models, a uses to predict who will fill in the application form (usage the decision tree method), the is the reputation valuation model. (Usage the nerve network method)From leave of 950,000 personal medium samples 700,000 samples again, use the model to find out which persons will make a reaction towards promoting sales the activity, and have the good reputation. Result is as follows: Include setting up the model use of 50,000 sampled 750,000 samples totally; an applicant is accepting, accepting the rate to rise 1.2% from 1%. The data mining although can't identify accurately which 10,000 applicants end will become the customer, can urge the marketing activity more valid. (2) Cross the value that the sale raises the existing customer through data mining. The product of the Guns and Rouses (G& R) Company sale is: Imitate the outdoors flower pot of the trench mortar and big gun and imitate the big caliber revolver with indoor flower pot of the long gun. The product watch is disheveled hair to go toward 12,000,000 families. When customer's telephone orders a certain product, (G& R) the company will promote other products actively- cross the sale. 43 学院理学学士论文 附录一 But,( G& R) the company detection only have 1/3 of customer allow they put forward the suggestion, at last cross the sale rate not enough 1%, and invited the one to complain the voice. Think to be certain for this B& R Company exactly is which persons need other products at the time of ordering a certain product. The G& R Company built up two types model of data mining, the one is used to predict whether a certain customer will is suggest to infuriate or not, the another is used to predict what kind of suggestion would accept well. The data mining model uses the customer the information database in the customer's information and new customer's information, tell the sales representative which grow the person to can adopt way and suggestions of cross the sale what product. The success rate that crosses the sale rose 2%, and little complaint. Data mining is to help the B& R Company better understood the customer's need. When cross the sales activity knot to the CRM that data mining model and typical model to put together, completely change of the general condition of the B& R Company. (3) Improve the value of customer through data mining. The Big Sam's Clothing is a dress company, which built up a website, providing the customer debarkation interface of the amity; carry on the on-line sale. Once you ordered a certain merchandise or register for the member, the company will recommend some your possible interested in new product Fs to be you toward you in the way of E-mail to check a product on the website, the website will recommend some your possible interested in product for having something to do with that products toward you. When the company builds up this website at the beginning, it did not consider all characteristics, and to be an on-line version of catalogue only. Ask the data mining tool for help the company improved the structure of the website, use gather a method and will press the product type classification to substitute before press to repair the principle classification with each other. When the customer looks for certain merchandise can use the set of these cents to provide the suggestion then, then build up customer's data, identify which customers the possibility is interested 44 学院理学学士论文 附录一 in knowing new product increase in the catalogue. At the same time, the Big Sam' the company of s predicts those new products that would cause the interest in the customer with the data excavation, and deliver through an E-mail to the customer, the customer can choose whether receive. The Big Sam's company made sure the characteristic strategy, passing to provide active sex customer the service guides the customer to choose the merchandise, not only increasing to sell the sum but also strengthening with the contact between customer. The customer has already seen it to adviser of make the allegiance of shopping. (4) Reserve the faithful customer through data mining. The company of Know Service is a network service company, which being like other companies also is at run off the customer without intermission. The rate of running off is monthly 8%, this means that 1,000,000 existing customers win to there will be monthly 80,000 customers run off. The cost that the company of Know Service re- looks for a customer is$200, the company want to invest in looking for the new customer monthly$16,000,000, therefore the company of Know Service demand the data mining for help. The company of Know Service controlled the customer in great quantities get to the Internet personal information of information and customer, set up customer's information database. The first matter that the Know Service Company does is from customer's database to win election to choose, change data. Connect down make according to the customer to expend, the life cycle carries on the demarcation to the customer, and judge the customer's last long, identify which customers will become the faithful customer then identify in faithful customer in future which run off probably. The third step, certain superior service, keeps the customer faithful. For example to get to the Internet to expend the service that big customer can provide to hire by the month but is not to provide more and greatly main page space. At result, the rate of running off of the customer is descending to 7.5% from original 8%, saving monthly$1,000,000. 45 学院理学学士论文 附录一 附 录 二 操作手册 (1)首先,进入登入界面,根据不同的部门,输入相应的用户名与密码,进入相应的 权限限制。 a) 进入用户信息管理模块,可以对用户信息进行浏览,可以增加 或删除用户,修改密码。 b) 进入客户资料模块,可以选择查看客户基本信息、判断客户属 性资料(粗糙集数据)、发掘潜在客户资料(关联规则数据)和分 析客户流失资料(决策树数据)。 c) 若选择查看“客户基本信息” d) 可以对客户信息进行浏览,可以增加、修改、删除,还可以按 不同的方式进行查询。 e) 若选择查看判断客户属性资料、发掘潜在客户资料和分析客户 流失资料,可以查看各个对应的数据表,进行增加或删除。 (2)进入客户分析模块,可以选择判断客户价值(粗糙集)、发掘潜在客户(关联规则)、 分析客户流失(决策树)和客户价值优化(遗传算法)。 1) 若选择“判断客户价值” a) 在弹出的对话框中单击“属性约简”,在下面的列表框中会输出 约简后的核属性和相对约简簇 b) 然后单击“值约简”,在下面的列表框中也会输出经过值约简后 的信息表。 c) 选择一组属性,就能判断这个客户是高价值的还是低价值的。 d) “显示信息”可以将信息显示到列表框中,同时我们还可以对列 表框中的信息进行“删除信息”和“删除所有信息”等操作。 2) 若选择“发掘潜在客户” a) 首先,选择单击“参数设置”,弹出参数设置对话框,其中“产 品总数”表示公司的产品个数,是一个大于0的数,“支持度” 是一个大于等于0小于等于1的数,表示产品被购买的频率,“置 信度”也是一个从0到1的数。设置好后点击OK。 b) 单击“频繁项目集”,输出频繁项目 c) 单击“输出强关联规则”,输出强关联规则到下面的列表框 d) 点击“输出潜在客户”得到我们所需要的客户 e) 如果想进一步了解潜在客户信息,可以按客户ID查询客户信息。 3) 若选择客户流失分析 a) 在弹出的对话框中选择单击“决策树导出”在下面的列表框中 会输出决策树流失模型,单击属性前面的“+”可以打开树的分 支 b) 这样我们选择下面下拉框中的一组属性,单击“判断是否流失” 就能判断这个客户是否存在流失的可能或是无法根据属性判 断。 46 学院理学学士论文 附录一 c) “添加用户信息”可以将选择属性信息显示到列表框中,同时 我们还可以对列表框中的信息进行“删除用户信息”和“清空 用户信息”的操作。 4) 若选择客户价值优化 a) 由于需要优化的函数和各参数已经经过默认输入,所以只要在弹 出的对话框中单击确定即可。 b) 单击后,会弹出另一对话框,表示要输入的染色体长度和变量 的上下限。 c) 输入后,单击,开始计算程序就会在下一个自动弹出的对话框 中显示结果,其中结果中的“染色体 0”表示客户关怀频率,“染 色体 1”表示商业交往频率。 47 学院理学学士论文 附录一 observing time of the whole fucking trip," Gaby muttered. In the seat next to Cirocco, Bill laughed. "Five minutes, Gaby," he said. "And I really think we ought to let them burn as long as we planned. It would work out so much nicer that way." The engines shut down on the tick and they watched for final confirmation that they were where they wanted to be. "This is Ringmaster; C. bones commanding. We have arrived in Saturn orbit at 1341.453 hours, Universal Time. I'll send up the prelims for a correcting burn when we come out from behind. Meanwhile, I'm going off this channel." She slapped the appropriate switch. "Anybody who wants to take a look outside, this is going to be your only chance." It was tight, but August and April and Gene and Calvin man- aged to squeeze into the cramped room. After checking with Gaby, Cirocco turned the ship ninety degrees. Saturn was a dark gray hole, seventeen degrees wide, covering 1000 times the area of the moon as seen from Earth. The rings were an incredible forty degrees from side to side. They looked like solid, brilliant metal. Ringmaster had come in north of the equator, so the upper face was presented to them. Each particle was being lit from the opposite side, presenting a thin crescent, like Saturn. The sun was a brilliant point of light in the ten o'clock position, approaching Saturn. No one spoke as the sun drew nearer to eclipse. They saw Saturn's shadow fall across the part of the ring nearest them, cutting it like a razor. Sunset lasted fifteen seconds. The colors were deep and changed rapidly, pure reds and yellows and blue-blacks like those seen from an airliner in the stratosphere. There was a soft chorus of sighs in the cabin. The glass depolarized and everyone gasped again as the rings grew brighter, bracketing the deep blue glow that outlined the northern hemi- sphere. Gray striations became visible on the planetary surface, illuminated by ringlight. Down there were storms as big as the Earth. When she looked away at last, Cirocco saw the screen to her left. Gaby was still in SCIMOD. There was an image of Saturn on the screen above her head, but she didn't look up at it. "Gaby, don't you want to come up and see this?" Cirocco saw her shake her head. She scanned the numbers marching across a tiny screen. "And lose the best observing time of the whole trip? You've got to he out of your mind." They first assumed a long, elliptical orbit with a low point 200 kilometers above the theoretical radius of Themis. It was a mathematical abstraction because the orbit was tilted thirty degrees from Themis, equator, which put them above the dark side. They passed the spinning toroid to emerge on the sun side. Themis lay spread out before them as a naked-eye object. Not that there was a lot to see. Themis was nearly as black as space, even with the sun shining on it. She studied the huge mass of the wheel with the triangular solar absorption sails rimming it like sharp gear teeth, presumably soaking up sunlight and turning it into heat. The ship moved over the interior of the great wheel. The spokes became visible, and the solar reflectors. They seemed nearly as dark as the rest of Themis, except where they mirrored some of the brighter stars. The problem that still worried Cirocco was the lack of an entrance. There was a lot of pressure from Earth to get into the thing, and Cirocco, despite her cautious instincts, wanted to as badly as anyone else. There had to be a way. No one doubted Themis was an artifaI think-" "The docking facilities!" Cirocco yelled. "They're gonna grab us! Bill, start the engine sequence, stop the carousel, get ready to move." "But it'll take us thirty minutes-" "I know. Move!" She caromed off the viewport and into her seat, reached for her microphone. "All hands. Emergency status. Depressurization alert. Evacuate the carousel. Acceleration stations. Strap in." She slammed the alarm button with her left hand and heard the eerie hooting begin in the room behind her. She glanced to her left. "You too, Bill. Get suited." "But-" "Now!" He was out of his seat and diving through the access hatch. She turned and called over her shoulder. "Bring my suit back with you!" The object was visible out the window now, approaching fast. She had never felt so helpless. By overriding the attitude control system's programing she was able to fire all the thrusters on the side of the ship facing Themis, but it was not nearly enough. The great mass of Ringmaster barely moved. Other than that, she could only sit and monitor the automatic engine sequencing and count the seconds as they dragged by. In a short time she knew they could not escape. That thing was big, and moving faster. Bill appeared, suited, and she scrambled into SCIMOD to don her own suit. Five anonymous figures sat belted to acceleration couches, not moving, staring at the screen. She clamped her helmet, and heard chaos. "Quiet down." The chatter died away. "I want silence on the suit channel unless 1 ask you to speak." "But what's happening, Commander?" It was Calvin's voice. 48 学院理学学士论文 附录一 "I said no talking. It looks like an automatic device is going to pick us up. This must he the docking facilities we were looking for." "It looks more like an attack to me," August muttered. "They must have done this before. They must know how to do it safely." She wished she could convince herself of that. it didn't help her credibility when the whole ship shuddered. "Contact," Bill said. "It's got us." Cirocco hurried back to her station, just in time to miss seeing the grapple sweep over them. The ship jumped again, and awful noises came from the rear. "What did it look like?" "Great big octopus tentacles without the suckers." He sounded shaken. "Tthere were hundreds of them, waving around all over." The ship gave an even greater lurch, and more alarms began to sound. A firestorm of red lights spread across her controls. "We've got a hull rupture," Cirocco said, with a calmness she did not feel. "Losing air from the central stem. Sealing off pres- sure doors 14 and IS." He hands moved over the controls with- out conscious guidance. The lights and buttons were far away, seen through the wrong end of a telescope. The accelerometer dial began to spin as she was thrown violently forward, then to the side. She came to rest on top of Bill, then struggled back to her scat and strapped in. When the buckle clicked around her waist the ship jerked backwards again, worse than before. Something came through the hatch behind her and hit the viewport, which developed a network of cracks. She hung from her seat, her body straining forward against the belt. An oxygen cylinder flew through the hatch. The glass shattered and the sound of the impact was sucked away with the burst of cold, hard glass knives that turned and dwindled before her eyes. Everything in the cabin that wasn't tied down leaped up and hurtled through the mouth of jagged teeth that had once been a viewport. Blood pulsed in her face as she hung above a bottomless black hole. Large objects turned lazily in the sunlight. One of them was the engine module of Ringmaster, out there in front of her where it had no right to be. She could see the broken stump of the connecting stem. Her ship was coming apart. "Oh, shit," she said, then had a vivid recollection of a tape she had once beard from the flight recorder of an airliner. That had been the last word the pilot had uttered, seconds before impact, when he knew he was going to die. She knew it, too, and the thought filled her with a vast disgust. She watched in dull horror as the thing that had the engines wrapped more tentacles around it. it reminded her of a Portuguese man o' way with a fish snared in its poisonous grip. A fuel tank ruptured---soundlessly, with a strange beauty. Her world was coming apart with no noise to mark its passing. A cloud of compressed gas quickly dispersed. The thing did not seem to mind. Other tentacles had other parts of the ship. The high-gain antenna almost seemed to be swimming away, but it moved too slowly as it tumbled down the well below her. "Alive," she whispered. "It's alive." "What did you say?" Bill was trying to hold himself secure with both hands m the instrument panel. He was strapped solidly to his chair, but the bolts which held it to the floor had broken. The ship shuddered again, and Cirocco's chair came free. The edge of the panel caught her across the thighs and she cried out as she struggled to free herself. "Rocky, things are falling apart in here." She wasn't sure whose voice it was, but the fear reached her. She pushed, and managed to open her seat belt with one hand while holding her- self away from the panel with the other. She slipped out to the side and saw her chair bounce across the shattered array of dials, stick briefly in the frame of the broken port, and launch into space. She thought her legs were broken, but found she could move them. The pain lessened as she drew on reserves of strength to help Bill out of his chair. Too late, she saw that his eyes were closed, his forehead and the inside of his helmet smeared with blood. As his body slithered loosely over the control panel she saw the dent his helmet had made in it. She fought for a grip on his thigh, then his calf, his booted foot, and he was falling, falling in the middle of a glittering shower of glass. She came to her senses crouched in the leg well under the control panel. She shook her head, unable to recall what had put her there. But the force of deceleration was not so great now. Themis had succeeded in bringing Ringmaster---or what was left of it-up to its own rotational speed. No one was talking. A hurricane of breathing came through the speaker in her helmet, but no words. There was nothing to say; the screams and curses had exhausted themselves. She got to her feet, grabbed the edge of the hatchway above her, and pulled herself through into chaos. No lights worked, but sunlight spilled harshly across broken equipment from a large rip in the wall. Cirocco moved through the debris and a suited figure got out of her way. Her head throbbed. One of her eyes was swollen shut. There was a lot of damage. It would take a while to get it cleaned up so they could get under weigh. th nobody'd ever want. me; Mum wouldn't say. anything, but she'd be cross after, and they used to come and visit, and so I kept a bottle in my room and I used to put it on. Once I drank some." Nicholas was silent. "Aren't you going to ask me if I died?" She stepped ahead of him, pulling aside the dripping branches. "See here, I'm sorry I said you were stupid." "I'm just thinking," Nicholas said. "I'm not mad at you. Do you really know anything about him?" "No, but look at it." She gestured. "Look around you; someone built all this." "You mean it cost a lot." "It's automated, of course, but still . " . well, the other places where you were before-how much space was there for each patient? Take the total volume and divide it by the number of people there." "Okay, this is a whole lot bigger, but maybe they think we're worth it." 49 学院理学学士论文 附录一 "Nicholas . . ." She paused. "Nicholas, Ignacio is homicidal. Didn't Dr. Island tell you?" "Yes." "And you're fourteen and not very big for it, and I'm a girl. Who are they worried about?" The look on Nicholas's face startled her. After an hour or more of walking they came to it. It was a band of withered vegetation, brown and black and tumbling, and as straight as if it had been drawn with a ruler. "I was afraid it wasn't going to be here," Diane said. "It moves around whenever there's a storm. It might not have been in our sector any more at all." Nicholas asked. "What is it?" "The Focus. It's been all over, but mostly the plants grow back quickly when it's gone." "It smells funny-like the kitchen in a place where they wanted me to work in the kitchen once." "Vegetables rotting, that's what that is. What did you do?" "Nothing-put detergent in the stuff they were cooking. What makes this?" "The Bright Spot. See, when it's just about overhead the curve of the sky and the water up there make a lens. It isn't a very good lens-a lot of the light scatters. But enough is focused to do this. It wouldn't fry us if it came past right now, if that's what you're wondering, because it's not that hot. I've stood right in it, but you want to get out in a minute." "I thought it was going to be about seeing ourselves down the beach." Diane seated herself on the trunk of a fallen tree. "It was, really. The last time I was here it was further from the water, and I suppose it had been there a long time, because it had cleared out a lot of the dead stuff. The sides of the sector are nearer here, you see; the whole sector narrows down like a piece of pie. So you could look down the Focus either way and see yourself nearer than you could on the beach. It was almost as if you were in a big, big room, with a looking-glass on each wall, or as if you could stand behind yourself. I thought you might like it." "I'm going to try it here," Nicholas announced, and he clambered up one of the dead trees while the girl waited below, but the dry limbs creaked and snapped beneath his feet, and he could not get high enough to see himself in either direction. When he dropped to the ground beside her again, he said, "There's nothing to eat here either, is there?" "I haven't found anything." "They-I mean, Dr. Island wouldn't just let us starve, would he?" - "I don't think he could do anything; that's the way this place is built. Sometimes you find things, and I've tried to catch fish, but I never could. A couple of times Ignacio gave me part of what he had, though; he's good at it. I bet you think I'm skinny, don't you? But I was a lot fatter when I came here." "What are we going to do now?" "Keep walking, I suppose, Nicholas. Maybe go back to the water." "Do you think we'll find anything?" From a decaying log, insect stridulations called, "Wait." Nicholas asked, "Do you know where anything is?" "Something for you to eat? Not at present. But I can show you something much more interesting, not far from here, than this clutter of dying trees. Would you like to see it?" Diane said, "Don't go, Nicholas." "What is it?" "Diane, who calls this `the Focus,' calls what I wish to show you `the Point."' Nicholas asked Diane, "Why shouldn't I go?" "I'm not going. I went there once anyway." 50 学院理学学士论文 附录一 "I took her," Dr. Island said. "And I'll take you. I wouldn't take you if I didn't think it might help you." "I don't think Diane liked it." "Diane may not wish to be helped-help may be painful, and often people do not. But it is my business to help them if I can, whether or not they wish it." "Suppose I don't want to go?" "Then I cannot compel you; you know that. But you will be the only patient in this sector who has not seen it, Nicholas, as well as the youngest; both Diane and Ignacio have, and Ignacio goes there often." , "Is it dangerous?" "No. Are you afraid?" . Nicholas looked questioningly at Diane. "What is it? What will I see?", She had walked away while he was talking to Dr. Island, and was now sitting cross-legged on the ground about five meters from where Nicholas stood, staring at her hands. Nicholas repeated, "What will I see, Diane?" He did not think she would answer. She said, "A glass. A mirror." "Just a mirror?" "You know how I told you to climb the tree here? The Point is where the edges come together. You can see yourself-like on the beach-but closer." Nicholas was disappointed. "I've seen myself in mirrors lots of times." Dr. Island, whose voice was now in the sighing of the dead leaves, whispered, "Did you have a mirror in your room, Nicholas, before you came here?" "A steel one." "So that you could not break it?" "I guess so. I threw things at it sometimes, but it just got puckers in it." Remembering dimpled reflections, Nicholas laughed. "You can't break this one either." "It doesn't sound like it's worth going to see." "I think it is." "Diane, do you still think I shouldn't go?" 51 学院理学学士论文 附录一 There was no reply. The girl sat staring at the ground in front of her. Nicholas walked over to look at her and found a tear had washed a damp trail down each thin cheek, but she did not move when Manuel, leaning against the barrera, watching the bull, waved his hand and the gypsy ran out, trailing his cape. The bull, in full gallop, pivoted and charged the cape, his head down, his tail rising. The gypsy moved in a zigzag and as he passed, the bull caught sight of him and abandoned the cape to charge the man. The gyp sprinted and vaulted the red fence of the barrera as the bull struck it with his horns. He tossed into it twice with his horns, banging into the wood blindly. The critic of El Heraldo lit a cigarette and tossed the match at the bull, then wrote in his notebook, ?large and with enough horns to satisfy the cash customers, Campagnero showed a tendency to cut into the terrain of the bullfighters.〃 Manuel stepped out on the hard sand as the bull banged into the fence. Out of the corner of his eye he saw Zurito sitting the white horse close to the barrera, about a quarter of the way around the ring to the left. Manuel held the cape close in front of him, a fold in each hand, and shouted at the bull. ?Huh! Huh!〃 the bull turned, seemed to brace against the fence as he charged in a scramble, driving into the cape as Manuel side-stepped, pivoted on his heels with the charge of the bull, and swung the cape just ahead of the horns. At the end of the cape he was facing the bull again and held the cape in the same position close in front of his body, and pivoted again as the bull recharged. Each time, as he swung, the crowd shouted. Four times he swung with the bull, lifting the cape so it billowed full, and each time bringing the bull around to charge again. Then, at the end of the fifth swing, he held the cape against his hip and pivoted, so the cape swung out like a ballet dancerˇs skirt and wound the bull around himself like a belt, to step clear, leaving the bull facing Zurito on the white horse, come up and planted firm, the horse facing the bull, its ears forward, its lips nervous, Zurito, his hat over his eyes, leaning forward, the long pole sticking out before and behind in a sharp angle under his right arm, held halfway down, the triangular iron point facing the bull. El Heraldoˇs second-string critic, drawing on his cigarette, his eyes on the bull, wrote: ˉthe veteran Manolo designed a series of acceptable veronicas, ending in a very Belmontistic recorte that earned applause from the regulars, and we entered the tercio of the cavalry.〃 Zurito sat his horse, measuring the distance between the bull and the end of the pic. As he looked, the bull gathered himself together and charged, his eyes on the horseˇs chest. As he lowered his head to hook, Zurito sunk the point of the pic in the swelling hump of muscle above the bullˇs shoulder, leaned all his weight on the shaft, and with his left hand pulled the white horse into the air, front hoofs pawing, and swung him to the right as he pushed the bull under and through so that the horns passed safely under the horseˇs belly and the horse came down, quivering, the bullˇs tail brushing his chest as he charged the cape Hernandez offered him. Hernandez ran sideways, taking the bull out and away with the cape, toward the other picador. He fixed him with a swing of the cape, squarely facing the horse and rider, and stepped back. As the bull saw the horse he charged. The picadorˇs lance slid along his back, and as the shock of the charge lifted the horse, the picador was already half-way out of the saddle, lifting his right leg clear as he missed with the lance and falling to the left side to keep the horse between him and the bull. The horse, lifted and gored, crashed over with the bull driving into him, the picador gave a shove with his boots against the horse and lay clear, waiting to be lifted and hauled away and put on his feet. Manuel let the bull drive into the fallen horse, he was in no hurry, the picador was safe; besides, it did a picador like that good to worry. Heˇd stay on longer next time. Lousy pics! He looked across the sand at Zurito a little way out from the barrera, his horse rigid, waiting. ?Huh!〃 he called to the bull, ?Tomar!〃 holding the cape in both hands so it would catch his eye. The bull detached himself from the horse and charged the cape, and Manuel, running sideways and holding the cape spread wide, stopped, swung on his heels, and brought the bull sharply around facing Zurito. ?Campagnero accepted a pair of varas for the death of one rosinante, with Hernandez and Manolo at the quites,〃 El Heraldoˇs critic wrote. ?He pressed on the iron and clearly showed he was no horse-lover. The veteran Zurito resurrected some of his old stuff with the pike-pole, notably the suerte ?Ol?! Ol?!〃 the man sitting beside him shouted. The shout was lost in the roar of the crowd, and he slapped the critic on the back. The critic looked up to see Zurito, directly below him, leaning far out over his horse, the length of the pic rising in a sharp angle under his armpit, holding the pic almost by the point, bearing down with all his weight, holding the bull off, the bull pushing and driving to get at the horse, and Zurito, far out, on top of him, holding him, holding him, and slowly pivoting the horse against the pressure, so that at last he was clear. Zurito felt the moment when the horse was clear and the bull could come past, and relaxed the absolute steel lock of his resistance, and the triangular steel point of the pic ripped in the bullˇs hump of shoulder muscle as he tore loose to find Hernandezˇs cape before his muzzle. He charged blindly into the cape and the boy took him out into the open arena. Zurito sat patting his horse and looking at the bull charging the cape that Hernandez swung for the shine of the pumping flow of blood, smooth against the black of the bullˇs shoulder. ?I got him that time,〃 Zurito said. Ch would be successful in keeping him away. Unwilling as she was to admit an idea so destructive of Jane's happiness, and so dishonourable to the stability of her lover, she could not prevent its frequently recurring. The united efforts of his two unfeeling sisters and of his overpowering friend, assisted by the attractions of Miss Darcy and the amusements of London, might be too much, she feared, for the strength of his attachment. As for Jane, her anxiety under this suspence was, of course, more painful than Elizabeth's; but whatever she felt she was desirous of concealing, and between herself and Elizabeth, therefore, the subject was never alluded to. But as no such delicacy restrained her mother, an hour seldom passed in which she did not talk of Bingley, express her impatience for his arrival, or even require Jane to confess that if he did not come back, she should think herself very ill used. It needed all Jane's steady mildness to bear these attacks with tolerable tranquillity. 52 学院理学学士论文 附录一 Mr. Collins returned most punctually on the Monday fortnight, but his reception at Longbourn was not quite so gracious as it had been on his first introduction. He was too happy, however, to need much attention; and luckily for the others, the business of love-making relieved them from a great deal of his company. The chief of every day was spent by him at Lucas Lodge, and he sometimes returned to Longbourn only in time to make an apology for his absence before the family went to bed. Mrs. Bennet was really in a most pitiable state. The very mention of any thing concerning the match threw her into an agony of ill humour, and wherever she went she was sure of hearing it talked of. The sight of Miss Lucas was odious to her. As her successor in that house, she regarded her with jealous abhorrence. Whenever Charlotte came to see them she concluded her to be anticipating the hour of possession; and whenever she spoke in a low voice to Mr. Collins, was convinced that they were talking of the Longbourn estate, and resolving to turn herself and her daughters out of the house as soon as Mr. Bennet were dead. She complained bitterly of all this to her husband. ``Indeed, Mr. Bennet,'' said she, ``it is very hard to think that Charlotte Lucas should ever be mistress of this house, that I should be forced to make way for her, and live to see her take my place in it!'' ``My dear, do not give way to such gloomy thoughts. Let us hope for better things. Let us flatter ourselves that I may be the survivor.'' This was not very consoling to Mrs. Bennet, and, therefore, instead of making any answer, she 53
本文档为【数据挖掘在客户关系管理中的—免费毕业设计论文】,请使用软件OFFICE或WPS软件打开。作品中的文字与图均可以修改和编辑, 图片更改请在作品中右键图片并更换,文字修改请直接点击文字进行修改,也可以新增和删除文档中的内容。
该文档来自用户分享,如有侵权行为请发邮件ishare@vip.sina.com联系网站客服,我们会及时删除。
[版权声明] 本站所有资料为用户分享产生,若发现您的权利被侵害,请联系客服邮件isharekefu@iask.cn,我们尽快处理。
本作品所展示的图片、画像、字体、音乐的版权可能需版权方额外授权,请谨慎使用。
网站提供的党政主题相关内容(国旗、国徽、党徽..)目的在于配合国家政策宣传,仅限个人学习分享使用,禁止用于任何广告和商用目的。
下载需要: 免费 已有0 人下载
最新资料
资料动态
专题动态
is_633808
暂无简介~
格式:doc
大小:534KB
软件:Word
页数:0
分类:互联网
上传时间:2017-10-08
浏览量:6