粗糙集理论的研究及其在电力业务数据挖掘中的应用

【作者】 徐敏；

【导师】 魏祖宽；

【作者基本信息】 电子科技大学 ， 计算机软件与理论， 2010， 硕士

【摘要】 RS理论是上世纪八十年代初由波兰数学家Pawlak提出的一种处理不精确知识的数学理论。其主要思想是利用已知的知识或信息来近似不精确的概念或现象。为快速高效地对海量GIS数据进行知识约简,分析了现有常见的约简算法及各自适用的范围。在此基础上,提出了新的基于遗传算法的约简方法,通过选取有效的适应度函数,很大程度上提高了对于大规模数据集的约简性能。并引入了自适应遗传算法的思想,使适应度函数中的因子能够随适应度的改变自动改变,从而增强了算法的全局寻优能力。通过对属性数据的实例分析,证明了该算法在海量数据约简方面的可行性和有效性。VPRS理论则是对传统粗糙集理论的一种扩充,它通过设置阈值β,放松了对标准粗糙集理论近似边界的严格定义,增强了粗糙集模型的抗干扰能力,噪声阈值β的不同,将对约简结果产生不一样的影响,因此要在实际应用的时候合理选择。针对传统RS理论不能完全处理不完备信息系统的瓶颈问题,将经典RS理论进行了不同的扩充,促进了其在实用化方向的发展。RS与其他技术的相互融合是今后发展的一个方向,能够有效地解决RS应用过程中遇到的一些瓶颈问题。如近几年发展迅猛的云模型、粒计算以及概念格等,作者重点关注的是它们在解决GIS-SDM方面的成就。通过在电网可视化管理系统(GVMS)平台中引入粗糙集理论的思想,可以在很大程度上增强系统的决策分析能力和电网的智能化管理。更多还原

【Abstract】 Rough set theory which was proposed by Pawlak in the early 1980s is a new mathematical theory to deal with ambiguous and uncertain knowledge. The main idea of the theory is to export the question’s decision-making or classification rules through the knowledge reduction while preserving the consistency of classification. To reduce the massive data of GIS quickly and efficiently, the existing common attribute reduction algorithms and their respective scope of application are analyzed. And further more, a new reduction method based on genetic algorithm (GA) is presented and performance of this method is largely improved by choosing an effective fitness function. And it also introduces the idea of adaptive genetic algorithm to promise the factors of fitness function changed automatically with the value of fitness function, then it can increase the ability of global optimization. The feasibility and efficiency of the presented method is demonstrated by analyzing the examples of attribute data in power business.Variable precision rough set is an extension of traditional rough set theory; it relaxes the restricted defining of approximation boundary in standard rough set theory and improves the anti-interference ability and Prediction ability to the new data of rough set model by setting up threshold valueβ. If noise thresholdβis different, the reduction results will be different too. Therefore the valueβshould be choice reasonably on the application. In view of the bottlenecks that traditional RS theory can not fully deal with incomplete information system, an expansion for traditional RS theory has been raised and contributed to the direction of RS practical development. A mutual confluence of RS and other technologies that concludes the cloud model, granular computing, and concept lattices, etc is a direction for future development that can solve a number of bottlenecks RS encountered effectively. The author will focus on their achievements in the settling the GIS-SDM.The introducing of rough set theory in the Grid Visualization Management System (GVMS) platform could enhance the decision analysis capabilities and the intelligent management of network in a large part.更多还原