大数据的分层分类优化识别仿真研究  

Simulation of Optimum Identification in Hierarchical Classification of Big Data

作　　者：张慧萍[1] 贺红艳[1] 陈小娟[1] 

机构地区：[1]湖北工业大学工程技术学院,湖北武汉430068

出　　处：《计算机仿真》2015年第10期463-466,共4页Computer Simulation

摘　　要：在对大数据环境下的分层分类挖掘过程中,由于非结构化数据大量涌现,使得数据结构属性不能统一,难以用表结构表示。采用传统方法时,不仅在记录数据数值的还要存储数据的结构,从而增加了数据分类的难度,导致分类精确度低的问题。提出改进叶贝斯理论的大数据环境下的分层分类挖掘方法。上述方法引入叶贝斯理论对数据库的数据进行详细的分析,输入大数据环境下的数据训练样本集,依据大数据训练样本集中每个数据的特征向量组建大数据的分层分类决策模型。在此基础上,利用最大间隔准则将分层分类模型中的每层高维数据投影到低维特征数据分类范围内,利用最小最大概率机对大数据进行分类优化。仿真证明,改进叶贝斯理论的大数据环境下的分层分类挖掘方法精确度高,适用性强。The paper proposed a hierarchical classification mining method under the big data environment based on improved Bayesian theory. In this method, Bayesian theory was introduced to make detailed analysis of the data in database, and the data training sample set under the big data environment was input. On the basis of the feature vector of each data in big data training sample set, the hierarchical classification decision model of big data was constructed. On this basis, the rule of maximum interval was used to project the high - dimensional data of each layer in the hierarchical classification model to the classification range of low - dimensional characteristic data, and minimum -maximum probability machine was used to make classification optimization of big data. Simulation experiment proves that the hierarchical classification mining method has high precision and strong applicability.

关 键 词：分类规则 数据挖掘 贝叶斯理论 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]