【作者】 王晶；

【导师】 姜丽红； 蔡鸿明；

【作者基本信息】 上海交通大学 ， 软件工程， 2009， 硕士

【摘要】 通信管理是重要的政府职能。通信管理部门需要掌握辖区内通信行业的业务分布与运营状况,来提高管理水平和决策水平。于是,及时准确的统计分析成为通信管理的重要内容。但业务种类繁多使得统计分析非常复杂。因此,建立电子政务系统,并进行网页类型的自动识别,从而帮助分类统计分析,对于通信管理具有较重要的意义。本论文进行了面向通信管理的电子政务系统的需求分析,完成了系统的设计与实现,并重点研究了系统中对抓取到的网页进行类型识别的网页分类算法。本论文进行的研究工作如下。1.面向通信管理的电子政务系统的架构及设计。本论文采用标准的关系型数据库、J2EE平台及目录服务作为系统的运行环境,构建了基于B/S的电子政务系统;搭建起包含门户、应用系统层、应用开发支持平台及数据库的系统架构;设计了系统中权限控制、审批流控制、在线业务等模块及网页分类模块。2.面向通信管理的电子政务系统的网页分类核心算法研究。本论文提出了一种新的网页分类算法CUCS(Combined UC and SVM)。CUCS算法在网页分类过程中吸收UC(Unsupervised Clustering)和SVM(Support Vector Machine)的各自优势,并通过修剪训练集进一步提高了网页分类的速度和准确率。CUCS算法最大的特色是“二次修剪”,它采用修剪过的训练集形成SVM分类器,然后再与UC结合,而非直接利用原始的训练集。因此,CUCS以较低的时间耗费取得了较高的分类准确率。3.面向通信管理的电子政务系统的实现及验证。本论文采用Java语言实现了面向通信管理的电子政务系统中的主要功能模块,包括:权限控制、审批流控制、许可申请业务、备案业务、年检业务等模块及网页分类模块;验证了从业务受理开始到网页分类最终实现的整个系统流程。更多还原

【Abstract】 Communication administration is an important government function. In order to improve administrative level and make strategic decisions, it is necessary for administration section on communication to grasp the business distribution and operation status of commutation industry. Statistical analysis that is both timely and exact has become an important part in the business of communication administration. However, various operations result in complex statistic. As a result, E-Government system which also can automatically classify web pages is urgently required so as to provide convenience for statistical analysis.This thesis conducted requirement analysis for Communication Administration Oriented E-Government system, and then finished the designing and implementation for the system. The thesis also studied particularly on web page classification algorithm which can automatically classify those grabbed web pages. This thesis has finished the following researching tasks.1. System architecture and designing of communication administration oriented E-Government System. RDBMS, J2EE and directory service are used for the running environment of system. The E-Government system is based on B/S mode. System architecture includes portal, application layer, development supporting platform and data base. Web page classification module, as well as modules of authority control, approval control and online business, has been designed in the system.2. Kernel algorithm research of web page classification in communication administration oriented E-Government System. This paper presents CUCS(Combined UC and SVM), a new algorithm for web page classification. CUCS combines the advantages of UC(Unsupervised Clustering)and SVM(Support Vector Machine). The biggest characteristic of CUCS is‘2-time pruning’. CUCS trained classifier by pruned training set rather than the initial training set, and then combined with UC. As a result, CUCS has made web page classification more exact with higher speed.3. System implementation and validation of communication administration oriented E-Government System. Java programming language is used to implement primary modules which include web page classification module, modules of authority control, approval control, license application business, backup business and annual exam business. The whole system running flow, which is from business handling to web page classification, has been validated in the thesis.更多还原