数据挖掘与数据融合相结合的信息处理技术研究

【作者】 王璿；

【导师】 高社生；

【作者基本信息】 西北工业大学 ， 导航、制导与控制， 2006， 硕士

【摘要】 随着全球信息化的快速发展,数据挖掘与数据融合这两种作为处理海量数据、提取有用信息的高新技术倍受瞩目。这两种技术处理数据的原理各不相同,但在功能上相互补充,整合这两种技术能更有效地解决工程中的实际问题。数据挖掘的首要问题是寻找数据来源,将来自不同样本的数据利用数据融合技术有效地加以综合,再进行数据挖掘。数据融合需要在已知的模型上进行,而数据挖掘技术则可自动地建立模型。将这两种技术进行深层次的结合与渗透,可协同完成复杂数据处理工作。 本文对数据挖掘与数据融合这两种技术进行了理论研究与应用探讨,研究内容及成果如下: (1)针对数据融合系统难以获取先验信息,以及存在大量冗余数据的问题,提出了用粗糙集理论建立数据融合模型的方法。采用属性约简及规则约简的方法处理数据,剔除冗余信息,获取最简规则。根据粗糙集最简规则建立融合系统。通过粗糙集离散化算法将融合方法推广至连续域。仿真结果证明,该算法是切实可行的。 (2)将模糊理论与神经网络技术相结合应用于数据融合。研究了基于模糊神经网络的融合系统原理,建立了基于模糊神经网络的融合系统,给出了神经网络学习算法。仿真结果验证了算法的可行性。 (3)针对数据挖掘与数据融合技术在功能上互补的特点,研究了数据挖掘与数据融合的集成技术,完善了知识获取过程,提高了融合精度。并建立了一种新的集成系统,研究给出了该集成系统的原理以及工作过程。 (4)针对粗糙集与模糊神经网络各自的优缺点,提出了一种粗糙集——模糊神经网络模型。研究了利用粗糙集技术获取模糊规则的方法,通过模糊神经网络对融合规则进行推广,克服了单独使用一种方法建立融合系统时存在的缺陷。 (5)设计了一种基于粗糙集——模糊神经网络技术的数据挖掘与数据融合集成系统。研究给出了设计原理,提出了建立集成系统的具体步骤,以及该集成系统的算法流程图。研究了集成系统各部分的算法,构建了网络的拓扑结构,推导出了网络的学习公式。提出了一种基于粗糙集方法的初始隶属函数获取算法,仿真结果证明了该集成系统的可行性与有效性。更多还原

【Abstract】 With the rapid development of information globalization, data mining and data fusion receive great attention as two promising technologies to deal with huge volume of data. Although the principles of both are different, they can complement each other. Integrating both technologies can efficiently solve practical engineering problem. The chief problem of data mining is to search data. The fusion system could efficiently synthesize the data from different samples. Data fusion is applied on virtual model, while data mining technology can automatically establish the model. Therefore, Once highly coupled they will be helpful to accomplish complicated data analysis automatically with effect.The dissertation researches the theories of data mining and data fusion. The hybrid system’s basic model is presented. The main work and achievements are as follows:(1) In order to deal with redundancy data, a new data fusion model based on rough sets is presented. Applying reduction method, the fusion algorithm is extracted. This fusion model could deal with continuous data , by using the discretization algorithm . Simulation results its feasibility.(2) One kind of fuzzy neural network, which is used to fuse multi-source data, is proposed by combing the advantages of neural network and fuzzy inference. Learning algorithm is also presented. Simulation results its feasibility.(3) Data mining and data fusion have complementary functions. According to their characteristics, the theory of integrated data mining and data fusion system is proposed.(4) A new fuzzy neural network based on rough sets is presented. Rules are extracted from database using rough sets to initialize the structure of fuzzy neural network. The fuzzy neural network is trained with original data. By combining those advantages of the two theories, the result is more satisfactory.(5) One kind of integrated data mining and data fusion system’s model is designed with fuzzy neural network based on rough sets. A principle diagram and flow chart of the integrated system are presented. The model of the network is constructed. Algorithm of this model is also studied. The simulation results prove that the proposed model showsgood performance in the learning speed and accuracy.更多还原