邻域粗糙集及其基于邻域粗糙集的分类算法-应用数学专业论文.docxVIP

摘 摘 要 经典粗糙集理论是定义在等价关系的基础上，只能处理名义型数据，对数值 型数据必须离散化后才可以处理．由于实际应用中的数据往往是数值的，而且测 量时本身带有误差，这给直接应用带来不便．连续数据经过离散化处理可能会丢 失重要信息并且不同的离散化策略会影响最终的处理效果．因此，数值型信息系 统的属性约简是近年来粗糙集理论研究的热点之一． 本文通过引入邻域关系粗糙集模型来进行数值属性约简和分类器构造．首先 研究了经典粗糙集理论及其性质，在此基础上，给出了邻域关系粗糙集理论及其 相关性质：其次，利用邻域粗糙集中的相关性质和邻域关系矩阵的性质对文[323 中属性约简算法做了改进，并提出了一种基于决策表的新的快速属性约简算法： 然后，针对传统KNN分类方法在分类时只考虑最近邻的样本点信息，而未考虑该 样本点的近邻信息，提出了一种考虑近邻信息的基于邻域粗糙集的分类算法：进 一步的考虑到样本的某个属性的取值特别大时，分类时会将其他取值小的属性的 信息淹没，于是提出利用压缩映射将数值较大的某些属性压缩到一个合理范围， 给出了基于压缩映射的邻域粗糙集分类方法；最后，对各种分类方法进行了比较 研究，并通过实例验证本文分类算法能够快速高效地进行分类． 关键词：邻域粗糙集：属性约简：邻域关系矩阵：KNN分类 t .. AbstractThe Abstract The Classical Rough Set Theory,defined on the basis of the equivalence relation， Call only deal、析th nominal data．Numerical data must be discredited before they call be handled．As the data in the practical application usually numerical，and not SO accurate itself嘶m some errors in the procession of the measurement，which have caused inconvenience to the direct application．In this condition，Discrimination of numeric data will lead to the loss of some important information，and different treatments and strategies will also affect the final results．Therefore，the Numeric Attribute Reduction of information systems is one of the hottest issues on Rough Set Theory in recent years． This paper introduces the Neighborhood Rough Set model to the numerical Attribute Reduction and Classification Structures．Firstly,it has studied the classical Rough Set Theory and its properties，and given the Neighborhood Rough Set Theory and its related properties．Secondly,the use of Rough Sets related to neighborhood character and the nature of the neighborhood relation matrix is offered to improve the Attribute Reduction Algorithm in the paper[32]，and a new and fast Attribute Reduction Algorithm based on decision table is proposed．Then by analyzing the traditional KNN Classification Method for classifying the sample which only considers the nearest poiIlt of information，without the account of the sample points of the neighbor