...
  • 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>International journal of systems assurance engineering and management >MR-TP-QFPSO: map reduce two phases quantum fuzzy PSO for feature selection
【24h】

MR-TP-QFPSO: map reduce two phases quantum fuzzy PSO for feature selection

机译：MR-TP-QFPSO：映射约简两相量子模糊PSO用于特征选择

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Feature selection is the utmost requirement to deal with high dimensional datasets. Fuzzy logic and particle swarm optimization are the two very popular soft computing methods which have used for feature selection. In this paper different variants of PSO are summarized to explore the latest development in PSO. The survey has been grouped in three categories; structures based PSO variants, fuzzy logic-PSO hybrids and parallel PSO variants. On the basis of findings of survey, map reduce two phases quantum behaved fuzzy rule PSO (MR-TP-QFPSO) method has been proposed. Quantum is the smallest possible state of any matter. Therefore, in proposed method smallest state of any particle is trit, which is having three values 0, 1 and #. # is included to bring a state of uncertainty where, feature is considered neither accepted nor rejected. In first phase search, feature space is exhaustively explored. During exhaustive initial search (first phase), multiple subsets of features are selected using quantum behaved fuzzy rule PSO (QFPSO). From these multiple subsets, minimum most important features (lower bound features) and maximum range of selected features are selected (upper bound feature subset). In second phase, selected feature subspace (selected in first phase) has been exploited and finally merged with lower bound features. The entire two phases search is highly iterative and it is well known that map reduce frame work can accelerate any iterative task by parallel processing. Therefore, proposed two phases QFPSO (TP-QFPSO) is applied using map reduce (MR-TP-QFPSO). The analysis of proposed algorithm clearly shows that map reduce has decreased the processing time of serial TP-QFPSO algorithm. The MR-TP-QFPSO is compared with other feature selection methods. The results on bench marking datasets show that MR-TP-QFRPSO outperformed the other methods. The reduction in execution time is directly propositional to the number of cluster nodes used. Therefore, as number of nodes is increased execution time will decrease without affecting the performance.
机译：特征选择是处理高维数据集的最高要求。模糊逻辑和粒子群优化是用于特征选择的两种非常流行的软计算方法。本文总结了PSO的不同变体，以探索PSO的最新发展。该调查分为三类：基于结构的PSO变体，模糊逻辑-PSO混合体和并行PSO变体。根据调查结果，提出了映射归约两相量子行为模糊规则PSO（MR-TP-QFPSO）方法。量子是任何事物中最小的状态。因此，在提出的方法中，任何粒子的最小状态都是trit，具有三个值0、1和＃。包含＃会带来不确定状态，在此状态下，功能既不被接受也不被拒绝。在第一阶段搜索中，将详尽探索特征空间。在详尽的初始搜索（第一阶段）期间，使用量子行为模糊规则PSO（QFPSO）选择多个特征子集。从这些多个子集中，选择最小的最重要特征（下界特征）和最大范围的选定特征（上界特征子集）。在第二阶段中，已利用选定特征子空间（在第一阶段中选定），并最终与下限特征合并。整个两个阶段的搜索都是高度迭代的，众所周知，map减少框架工作可以通过并行处理来加速任何迭代任务。因此，建议的两个阶段QFPSO（TP-QFPSO）使用地图约简（MR-TP-QFPSO）应用。对该算法的分析清楚地表明，映射约简减少了串行TP-QFPSO算法的处理时间。将MR-TP-QFPSO与其他功能选择方法进行了比较。基准标记数据集的结果表明，MR-TP-QFRPSO优于其他方法。执行时间的减少直接取决于所使用群集节点的数量。因此，随着节点数量的增加，执行时间将减少而不会影响性能。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号