文档数据库结构信息分类筛查方法仿真
|
摘要
当前数据库信息分类筛查方法准确率较低、丢包率较高,提出基于PSO和DNN的文档数据库结构信息分类筛查方法。利用SNR特征基因选择法对文档数据库结构信息特征属性进行降维,采用OCDD算法对连续型的数据信息进行离散化处理。将数据库结构信息某类别标签和属性变量的互信息与信息熵之间比值当作信息离散化目标函数,依据动态迭代规划法对目标函数进行求解,得到最优离散化分区。在自动编码机顶层设计一个Softmax信息分类器,通过粒子群算法对自动编码机权值进行优化。Softmax信息分类器在编码结束后根据梯度下降法对自身进行调整,同时将分类器代价函数当作调整编码机权值过程中一个评估值,共同和编码机误差函数指导权值优化。利用优化后的权值和Softmax信息分类器完成文档数据库结构信息分类筛查。实验结果表明,该方法平均丢包率为0.28%,分类筛查准确性良好,具备可靠性能。
This article puts forward a method to classify and screen the document database structure information based on PSO and DNN. Firstly, we used SNR feature gene selection method to reduce the dimensionality of feature attribute of document database structure information, and then we used OCDD algorithm to discretize the continuous data information. Secondly, we regarded the ratio between the mutual information and the information entropy of a category label and attribute variable of database structure information and as the discretization objective function. Thirdly, we solved the objective function based on the dynamic iterative programming method to obtain the optimal discretization subarea. Moreover, we designed a Softmax information classifier on the top of automatic encoder and optimized the weight of automatic encoder by particle swarm optimization. After the end of encode, Softmax information classifier adjusted itself according to the gradient descent method. Meanwhile, we regarded the classifier cost function as an evaluation value in the process of adjusting the weight of encoder, which was used to guide the weight optimization with the error function of encoder. Finally, we used the optimized weight and Softmax information classifier to complete the classification and screening of document database structure information. Simulation results show that the average packet loss rate of the proposed method is 0.28%. Meanwhile, the accuracy of classification screening is good and reliable.
This article puts forward a method to classify and screen the document database structure information based on PSO and DNN. Firstly, we used SNR feature gene selection method to reduce the dimensionality of feature attribute of document database structure information, and then we used OCDD algorithm to discretize the continuous data information. Secondly, we regarded the ratio between the mutual information and the information entropy of a category label and attribute variable of database structure information and as the discretization objective function. Thirdly, we solved the objective function based on the dynamic iterative programming method to obtain the optimal discretization subarea. Moreover, we designed a Softmax information classifier on the top of automatic encoder and optimized the weight of automatic encoder by particle swarm optimization. After the end of encode, Softmax information classifier adjusted itself according to the gradient descent method. Meanwhile, we regarded the classifier cost function as an evaluation value in the process of adjusting the weight of encoder, which was used to guide the weight optimization with the error function of encoder. Finally, we used the optimized weight and Softmax information classifier to complete the classification and screening of document database structure information. Simulation results show that the average packet loss rate of the proposed method is 0.28%. Meanwhile, the accuracy of classification screening is good and reliable.
引文
[1] 林钰杰,吴丽贤.基于Lucene的企业电子文档搜索系统的开发研究[J].电子设计工程,2017,2(17):102-106.
[2] 赵一,等.微博演化网络的负信息分类方法[J].计算机科学与探索,2017,11(1):91-98.
[3] 杜永萍,刘京旋,张津丽.基于多语义元路径的异质网节点分类方法[J].模式识别与人工智能,2017,30(12):1100-1107.
[4] 赵英,谭杨.基于相关信息的网络流量贝叶斯分类法改进[J].计算机工程,2016,42(3):80-83.
[5] 吕淑宝,等.一种深度学习的信息文本分类算法[J].哈尔滨理工大学学报,2017,22(2):105-111.
[6] 王浩,等.一种基于多阶邻居的网络环境下多标签分类算法[J].电子学报,2016,44(10):2330-2334.
[7] 李辉,刘昊泽,牛怿.一种基于C4.5算法的广告数据包分类方法[J].北京化工大学学报:自然科学版,2016,43(5):101-106.
[8] 刘卓然,等.基于标签相似度的不良信息多标签分类方法[J].计算机应用研究,2016,33(4):989-992.
[9] 陈丁,赵军,吴春旺.互联网中混合入侵信息节点定位识别仿真[J].计算机仿真,2017,34(7):195-198.
[10] 杜慧,等.基于情感词向量的微博情感分类[J].中文信息学报,2017,31(3):170-176.
[2] 赵一,等.微博演化网络的负信息分类方法[J].计算机科学与探索,2017,11(1):91-98.
[3] 杜永萍,刘京旋,张津丽.基于多语义元路径的异质网节点分类方法[J].模式识别与人工智能,2017,30(12):1100-1107.
[4] 赵英,谭杨.基于相关信息的网络流量贝叶斯分类法改进[J].计算机工程,2016,42(3):80-83.
[5] 吕淑宝,等.一种深度学习的信息文本分类算法[J].哈尔滨理工大学学报,2017,22(2):105-111.
[6] 王浩,等.一种基于多阶邻居的网络环境下多标签分类算法[J].电子学报,2016,44(10):2330-2334.
[7] 李辉,刘昊泽,牛怿.一种基于C4.5算法的广告数据包分类方法[J].北京化工大学学报:自然科学版,2016,43(5):101-106.
[8] 刘卓然,等.基于标签相似度的不良信息多标签分类方法[J].计算机应用研究,2016,33(4):989-992.
[9] 陈丁,赵军,吴春旺.互联网中混合入侵信息节点定位识别仿真[J].计算机仿真,2017,34(7):195-198.
[10] 杜慧,等.基于情感词向量的微博情感分类[J].中文信息学报,2017,31(3):170-176.
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |