【作者】 郭翠翠；

【导师】 姜德生； 钟珞；

【作者基本信息】 武汉理工大学 ， 计算机应用技术， 2010， 博士

【摘要】 城市交通作为支撑经济发展的重要基础,城市中的各种交通基础设施修建的速度也越来越快。地下空间是公路隧道、过江隧道、地铁通道的统称,具有多样性以及复杂性。先期的隧道安全预测预警的主要监测目标是隧道交通的通畅程度,主要监测对象是隧道内交通事故、车辆拥挤程度、火灾等灾难发生等。在当前的社会背景与技术条件下,地下空间监控预警不仅具有早期的隧道交通的通畅程度监视功能,而且集地下空间结构安全检测与评估、在役环境监测与控制、综合信息采集与通信、数据传输与处理、安全评价与预测为一体,成为一个综合的地下空间监控预警系统。人们往往会把地下空间的信息按其各自特征和性能划分为若干较为简单的块以帮助人们更好地认识、理解它。每个被划分的块可以看成是一个粒,人们可以通过不同的粒度来更好地研究地下空间。本文探索了在不同信息粒度情况下对属性集进行属性约简,优化数据预处理的方法。为解决在海量数据、新增加数据条件下,进一步增进粒计算具体模型的学习能力、分类与预测能力、推广能力奠定基础。在研究粒的计算方面,研究了决策方法,讨论在粒度不同的情况下,粒的计算问题及决策挖掘。在完成上述研究的基础上,给出了基于rough布尔代数的粒计算刻画。对粒的运算进行了探讨性定义和描述,给出了rough布尔代数下的粒原子表示,并定义了粒格。最后,通过具体工程数据,对上述新的方法进行了可行性验证。(1)在深入介绍粒计算含义的基础上,对粒计算基本研究问题进行了分析。包括信息粒化、粒的构建与运算,以及粒层的概念和特性。随之比较了粗糙集、模糊集和商空间等常见的粒计算模型,分析了各个模型的特点及交叉研究方向。并且对信息系统的全粒度空间进行了描述。(2)首先从集合的观点与算子的观点这两种形式描述了粗糙集,然后分别从代数表示法和信息表示法出发给出了属性约简的算法。进而根据二进制粒构建的粒矩阵,提出了粒矩阵相与运算；根据相与运算的方法,描述了在信息守恒的前提下,基于粒矩阵的知识粒化方法；给出基于粒矩阵的启发式属性约简方法GrM-AR,为属性约简提供了一个新的思路,并且通过理论分析证明该算法是可行有效的。(3)根据二进制粒构建的粒矩阵,扩展并构建了标志粒矩阵,提出了标志粒矩阵相与运算,进而得到了标志分段粒矩阵,并且根据相与运算的方法提出了基于标志粒矩阵规则挖掘算法SGrM-DR,为规则挖掘提供了新思路。(4)基于rough布尔代数的粒计算刻画方法。描述了基于给定信息系统上定义的以rough布尔代数为平台的粒计算及其推理的详细刻画。分别描述了rough布尔代数下的上近似和下近似运算,并对具体的粒的运算给出了详细推理和证明。在rough布尔代数的原子表示基础上,提出了粒格的定义,并充分证明了定理的正确性。对rough布尔表示的属性约简研究奠定了一定的理论基础。(5)分别从数据处理及预警规则挖掘两方面阐述了基于粒计算的地下空间预警方法,并分别从理论和应用两方面分析了其可行性。通过部分样本数据示例,充分说明了前述研究的有效性与意义。为地下空间监控预警规则挖掘提供了新的途径。更多还原

【Abstract】 The transportation of a city is the basic foundation of economy development. The establishment of transportation infrastructure comes along with a high speed. Underground space is a general designation of city tunnels, cross-river tunnels and subways. There must be diversity characters and complex characters with it. As one of the most important part of underground space, the prediction and prewarning of city tunnels was highly focused on by many researchers. The prediction and prewarning of city tunnels in advance was mainly focused on the extent of transportation smooth, the objects are transportation incidents, congestion in tunnels, fire tragedies and so on. The technical measures are using cameras to connect transportation situation with supervise control systems. Under current society background and technical conditions, the underground prediction and prewarning monitoring systems are not only a transportation monitoring system, but also a combination system having abilities of safety examining and evaluating, information collection and communication, datum transferring and processing. People are accustomed to classify underground information into several blocks by their characters and performance, to help people easily and effectively understand the information. Each such block is acknowledged as a granular. Therefore, people can do research and application on underground space by such different granular.This paper was doing some research on attributes reduction in different information granularity, to optimize the method of pretreatment datum. For granular characters such as solving uncertain, incomplete, pare-true and large datum problems, it has been widely used in recent years. On the aspect of granularity theory, some researching work has been done on decision making method. After that, it proposes a kind of description of granular using rough bool algebraic theory. Finally, it uses engineering datum to validate the correctness of methods above by both attributes reduction algorithm and decision making algorithm.(1) With such more works on basic theory of granular, this paper researched and analyzed many problems existed in this field, such as information granulation, the construction and computation of granular, the conception and characters of granular layer. After that work, this paper made comparison with normal models of granular computing such as rough sets, fuzzy sets and quotient space. And this paper analyzed each model’s characters and combination research direction. Finally, we described the whole granular space of information system.(2) First, we described rough sets respectively from the aspect of aggregation and aspect of arithmetic operators, then proposed the attributes reduction algorithms respectively from the aspect of algebra expression and information expression. This paper constructed the granular matrix by binary granular, proposed the operation of granular matrix "AND". Based on this method, this paper described the method of knowledge granulation by the theory of information conservation. Then, this paper proposed the attributes reduction algorithm GrM-AR, a new way to make attributes reduction. It was evaluated executive and effective by theoretical analysis.(3) Based on granular matrix above, we made symbol granular matrix, and proposed symbol granular matrix "AND" operation, to get symbol subsection granular matrix. Then algorithm for rules extraction SGrM-DR was evaluated executive and effective by theoretical analysis.(4) This paper was doing description of granular using rough bool algebraic theory and proposed the operation details and proves procedures details. Based on above theoretical foundations, this paper proposed the definition of granular metre. It established much more basis on granular models based on rough bool description methods.(5) At last, a kind of prediction and prewarning method of underground was illustrated from the aspect of datum processing and rules extracting. It was executive by theoretical analysis and application analysis. By some concretely experiments of engineering datum, it was proved completely. It proposed a new idea of prediction and prewarning of underground space monitoring.更多还原