北京邮电大学学报

  • EI核心期刊

北京邮电大学学报 ›› 2007, Vol. 30 ›› Issue (6): 85-88.doi: 10.13190/jbupt.200706.85.yuyh

• 论文 • 上一篇    下一篇

改进的基于支持向量机的网络综合评价策略

于艳华,宋梅,潘阳发,宋俊德   

  1. (北京邮电大学 电子工程学院,北京 100876)
  • 收稿日期:2007-03-27 修回日期:2007-07-19 出版日期:2007-12-31 发布日期:2007-12-31
  • 通讯作者: 于艳华

An Improved Network Performance Evaluation Method Based on Support Vector Machines

YU Yan-hua, SONG Mei, PAN Yang-fa, SONG Jun-de   


  1. (School of Electronic Engineering, Beijing University of Posts and Telecommunications, Beijing 1000876, China)
  • Received:2007-03-27 Revised:2007-07-19 Online:2007-12-31 Published:2007-12-31
  • Contact: Yu Yan-Hua

摘要:

针对现有移动网络性能综合评估方法中存在的问题,提出了在维度变换基础上的采用支持向量机的综合评价策略。首先对语义上相关的n个指标进行维度变换使之成为独立的n维,然后对变换后的数据用支持向量机建立回归模型。理论分析表明,这种方法既可克服反向传播(BP)神经网络方法在应用中存在的收敛于局部极小问题,也可避免主成分分析法引起的信息丢失问题。实验表明,用支持向量机的方法比用BP神经网络的方法过程更可控,预测误差更小,且样本评价值间的差异保持得更好。

关键词: 反向传播神经网络, 主成分分析; 支持向量机; 维度变换

Abstract:

Evaluation of the performance of mobile network and its elements is the basis of network optimization. According to the problems existing in the applications of the methods applied at present, a new method based on dimension transformation and support vector machines was proposed. The steps were that, firstly, transforming the n related indicators to another n independent indicators, and secondly, using support vector machines (SVM) to model the transformed data. Theoretical analysis shows that this method can conquer the problems of back propagation(BP) neural network: overfitting,and the danger of getting stuck into local minima. The information loss occurring in the application of primary component analysis was avoided. Experimental results show that compared to BP neural network, the training process of support vector machines is more controllable, and the relative error of evaluation score based on support vector regression machines is smaller. Furthermore, the evaluation differences of the samples are maintained better.

Key words: back propagation neural network, primary component analysis, support vector machines, dimension transformation

中图分类号: