...
首页> 外文期刊>Journal of Residuals Science & Technology >Research on KD-BP Neural Network Feature Selection Optimization Algorithm for Big Data Sets
【24h】

Research on KD-BP Neural Network Feature Selection Optimization Algorithm for Big Data Sets

机译:大数据集的KD-BP神经网络特征选择优化算法研究

获取原文
   

获取外文期刊封面封底 >>

       

摘要

This paper studies and analyzes data from the perspective of data availability, and uses the KD-BP neural network optimization algorithm to determine and extract reliable data classification features for big data sets. The algorithm introduces the k-nearest neighbor search method to quickly eliminate the redundant features, ensuring the maximized reservation of the valid evidence of input nodes, to simplify the structure of the algorithm and to reduce the complexity of the algorithm. Introducing a self-adaptive weight impulse term model, it can avoid the premature convergence and easily falling into the situation of the local minimum point in the BP neural network. Finally, through the best feature weight selected by calculating the fusion degree among the nodes of each layer, we can achieve the optimal selection of the classification features. The comparison between the KD-BP algorithm and the standard BP algorithm proves that the KD-BP neural network algorithm is better than the standard BP algorithm in feature selection.
机译:本文从数据可用性的角度研究和分析数据,并使用KD-BP神经网络优化算法来确定和提取大数据集的可靠数据分类特征。该算法引入了k最近邻搜索方法,以快速消除冗余特征,确保最大程度地保留输入节点的有效证据,从而简化了算法的结构并降低了算法的复杂性。引入自适应权重脉冲项模型,可以避免过早收敛,容易陷入BP神经网络局部极小点的情况。最后,通过计算各层节点之间的融合度,通过选择最佳特征权重,可以实现分类特征的最优选择。通过对KD-BP算法和标准BP算法的比较证明,在特征选择方面,KD-BP神经网络算法优于标准BP算法。

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