【作者】 张瑞军；

【导师】 陈定方；

【作者基本信息】 武汉理工大学 ， 计算机应用技术， 2007， 博士

【摘要】 随着信息技术在企业的日益普及，信息系统在工具手段、开发模式、软件规模、指导思想等方面的不断提升，企业在信息需求方面日趋多样化、人性化、智能化，智能信息处理成为信息系统发展的一个重要方向。但是传统的智能信息处理方法只是简单停留在人工智能方法的引入，真正从数据→信息→知识这一信息链的角度来看，目前的智能信息处理方法有如下几个方面不足：第一、在数据的自动获取方面，提出了一些智能化的仪器仪表，但是针对多信号源情况下如何完成数据自动、准确、无冗余的提取方法没有太多的研究；第二、信息处理模式单一，对多种结构、多种来源信息的组织和呈现方式等方面研究不足；第三、在知识发现方面，对多领域专家情景下的知识发现涉足不多。本论文从信息链的角度，在大量查阅国内外文献的基础上，依托大型信息系统项目和科学研究课题，针对其中的关键技术，按照系统建模、形式化描述、算法研究、实验验证与对比这一技术路线，在智能数据采集、智能信息处理和知识发现方面做了一些研究，主要体现在以下几个方面：(1)针对无人值守情况下多源数据采集中的身份识别与自动称重等具有代表性的问题，在分析业务流程和通信协议的基础上，利用软件工具中定时器交互机制和循环队列中数据存储与访问的特点，提出了一种基于时隙和循环队列的智能数据采集算法，对时间块长度、队列规模等进行了研究。将该算法应用于武汉市城管局垃圾场垃圾计量项目中，取得了很大的成功。(2)从白盒法和黑盒法相结合的角度提出了一种构件的3C模型，设计了一种基于该模型的树形结构“超”实体，给出其构造及相关算法，并以树形结构“超”实体作为多层次实体的信息组织方式。同时在结合QQ式的菜单和1／K开关式构件连接方式的基础上，提出了一种基于可编辑电子书构件的智能信息呈现方法，很好地解决了结构化、半结构化、非结构化数据在同一信息系统平台呈现的问题，并在长江防洪工情信息服务系统和全国山洪防治规划信息系统中得到很好的应用。(3)对于多领域专家情景下的知识发现问题，利用BP神经网络学习能力强、具有较好鲁棒性等特点，在修改BP神经网络学习率等参数的基础上，针对复杂系统中网络学习的问题，提出了一种网络拓扑可重构的BP网络“超”结点，用于各子系统的知识组织与学习，并以专家系统为骨架，结合模糊逻辑，构造了基于BP可重构网络的协同模糊专家系统模型，并将该模型成功应用于青年学生素质评估。(4)对于装配线平衡(ALB)这类具有代表性的NP难问题中的知识发现，利用遗传算法强搜索能力的特点，引进惩罚因子、适应度定标、随机普遍取样等概念，ALB问题的编码规则和遗传算法中诸算子进行了改进，构建了基于改进遗传算法的ALB问题求解模型，对其关键技术进行了描述，并给出了算法运行实例和实验对比数据。更多还原

【Abstract】 With the popularization of information technology in enterprise and the increase of the following aspects of information system: development tools, development mode, software dimension, guiding ideology and so on, demands for information become more and more various, personal and intelligent, intelligence information processing has become an important development trend of information system. The traditional methods depend only on the introduction of artificial intelligence. There are some faults in intelligence information processing from the information chain viewpoint of data, information and knowledge. Firstly, in the aspect of data automatic acquisition, although several intelligent instruments have been brought forward, less research was made on how to acquire data automatically, correctly and with no redundancy under the condition of multiple signal sources. The Second is the simple pattern in information processing and the lack study of organization and presentation of multi-structure and multi-source information. Finally, knowledge discovery with multi-field and multi-expert has been less discussed.On the basis of consulting a great deal of domestic and foreign literatures, the paper has made some research on intelligent data acquisition, intelligence information processing and knowledge discovery from the viewpoint of information chain, which aimed at the key technologies of large information system project and scientific research task according to the technical route of system modeling, formal description, algorithm research, experiment validation and contrast. The major research results are the following:(1) An intelligent data acquisition algorithm based on time slot and circular queue is brought forward in allusion to the representative multi-source data acquisition such as identity recognition and auto-weight under unattended condition. By analyzing the system flow and data transportation protocols, the key parameters for the algorithm such as the length of time block, the size of queue and so on are discussed, which make use of the timer interaction principle in software tool and the character of data store and access in circular queue. Great success is achieved when the algorithm is used in the garbage measurement project for Wuhan city administrative bureau.(2) A 3C-model of component is brought through by combining the white box and black box method. A tree-structure super entity based on the model is designed and its constructional and correlative algorithm is also given, which acts as the information organization method of multiple level entities. An intelligent information presentation method based on editable e-book component is presented by combining QQ menu and 1/K switcher components-connection method, which solved soundly the problem how structured, half-structured and non-structured data can display in the same information system platform. It is widely used in the embankment information system of Yangtze River and nation-wide mountain torrents control programming information system.(3) BP neural network is robust and of strongly learning ability. By modifying the parameters such as learning rate of BP neural network and so on, the concept of BP super node with the network frame can be reconfigurable is presented, which aimed at knowledge discovery problem with several field experts. It is used in knowledge organization and learning of each subsystem. A cooperation fuzzy expert system model based on BP network is constructed, with a framework of expert system by combining fuzzy logic and BP super node. The model is successfully used in the diathesis evaluating of youth students.(4) The assembly line balancing(ALB)problem is a representational NP-hard problem. Genetic algorithm has strongly search ability. A model based on improved GA is designed to solve the knowledge discovery problem of ALB, in which the concept of penalty factor, scaling fitness and stochastic universal sampling are introduced. The coding rule of ALB and several operators of GA are improved in the model. The key technology is described for the model. The running example and experiment contrast data are also given.更多还原