基于深度信念网络和线性单分类SVM的高维异常检测

doi:10.11959/j.issn.1000-0801.2018006

电信科学 ›› 2018, Vol. 34 ›› Issue (1): 34-42.doi: 10.11959/j.issn.1000-0801.2018006

基于深度信念网络和线性单分类SVM的高维异常检测

李昊奇,应娜,郭春生,王金华

杭州电子科技大学，浙江杭州 310018

修回日期:2017-09-26 出版日期:2018-01-01 发布日期:2018-02-05
作者简介:李昊奇（1992-），男，杭州电子科技大学硕士生，主要研究方向为深度学习与数据挖掘。|应娜（1978?），女，博士，杭州电子科技大学副教授、硕士生导师，主要研究方向为信号处理与人工智能。|郭春生（1971?），男，博士，杭州电子科技大学副教授、硕士生导师，主要研究方向为模式识别与人工智能。|王金华（1992-），女，杭州电子科技大学硕士生，主要研究方向为深度学习与自然语言处理。
基金资助:
国家自然科学基金资助项目(61372157);“电子科学与技术”浙江省一流学科A类基金资助项目(GK178800207001)

High-dimensional outlier detection based on deep belief network and linear one-class SVM

Haoqi LI,Na YING,Chunsheng GUO,Jinhua WANG

Hangzhou Dianzi University,Hangzhou 310018,China

Revised:2017-09-26 Online:2018-01-01 Published:2018-02-05
Supported by:
The National Natural Science Foundation of China(61372157);Zhejiang Provincial First Class Disciplines:Class A-Electronic Science and Technology(GK178800207001)

摘要/Abstract

摘要：

针对目前高维数据异常检测存在的困难，提出一种基于深度信念网络和线性单分类支持向量机的高维异常检测算法。该算法首先利用深度信念网络具有良好的特征提取功能，实现高维数据的降维，然后基于线性核函数的单分类支持向量机实现异常检测。选取UCI机器学习库中的高维数据集进行实验，结果表明，该算法在检测正确率和计算复杂度上均有明显优势。与PCA-SVDD算法相比，检测正确率有4.65%的提升。与自动编码器算法相比，其训练和测试时间均有显著下降。

关键词: 异常检测, 高维数据, 深度信念网络, 单分类支持向量机

Abstract:

Aiming at the difficulties in high-dimensional outlier detection at present,an algorithm of high-dimensional outlier detection based on deep belief network and linear one-class SVM was proposed.The algorithm firstly used the deep belief network which had a good performance in the feature extraction to realize the dimensionality reduction of high-dimensional data,and then the outlier detection was achieved based on a one-class SVM with the linear kernel function.High-dimensional data sets in UCI machine learning repository were selected to experiment,result shows that the algorithm has obvious advantages in detection accuracy and computational complexity.Compared with the PCA-SVDD algorithm,the detection accuracy is improved by 4.65%.Compared with the automatic encoder algorithm,its training time and testing time decrease significantly.

Key words: outlier detection, high-dimensional data, deep belief network, one-class SVM

中图分类号:

TP183

李昊奇,应娜,郭春生,王金华. 基于深度信念网络和线性单分类SVM的高维异常检测[J]. 电信科学, 2018, 34(1): 34-42.

Haoqi LI,Na YING,Chunsheng GUO,Jinhua WANG. High-dimensional outlier detection based on deep belief network and linear one-class SVM[J]. Telecommunications Science, 2018, 34(1): 34-42.

图/表 11

图1

图2

表1

表2

图3

图4

表3

表4

表5

表6

图5

参考文献 20

[1]	王忠伟, 陈叶芳, 肖四友 ,等. 一种高维大数据全k近邻查询算法[J]. 电信科学, 2015,31(7): 52-62.
	WANG Z W , CHEN Y F , XIAO S Y ,et al. An AkNN algorithm for high-dimensional big data[J]. Telecommunications Science, 2015,31(7): 52-62.
[2]	CHANDOLA V , BANERJEE A , KUMAR V . Anomaly detection:A survey[J]. ACM Computing Surveys, 2009,41(3): 1-58.
[3]	SHIN H J , EOM D H , KIM S S . One-class support vector machines—an application in machine fault detection and classification[J]. Computers ＆ Industrial Engineering, 2005,48(2): 395-408.
[4]	李昕, 钱旭, 王自强 . 一种高效的高维异常数据挖掘算法[J]. 计算机工程, 2010,36(21): 34-36.
	LI X , QIAN X , WANG Z Q . Efficient data mining algorithm for high-dimensional outlier data[J]. Computer Engineering, 2010,36(21): 34-36.
[5]	TENENBAUM J B , DE S V , LANGFORD J C . A global geometric framework for nonlinear dimensionality reduction[J]. Science, 2000,290(5500): 2319.
[6]	POMERANTSEV A L . Principal component analysis(PCA)[M]. New York: John Wiley ＆ Sons,Inc., 2014: 4229-4233.
[7]	ROWEIS S T , SAUL L K . Nonlinear dimensionality reduction by locally linear embedding[J]. Science, 2000,290(5500): 2323.
[8]	GONZALEZ I,DéJEAN S , MARTIN P G P ,et al. CCA:an R package to extend canonical correlation analysis[J]. Journal of Statistical Software, 2008,23(12).
[9]	CHENOURI S , LIANG J , SMALL C G . Robust dimension reduction[J]. Wiley Interdisciplinary Reviews Computational Statistics, 2015,7(1): 63-69.
[10]	程辉, 方景龙, 王大全 ,等. 超平面支持向量机简化性能分析[J]. 电信科学, 2015,31(8): 78-83.
	CHENG H , FANG J L , WANG D Q ,et al. Performance analysis of simplification of hyperplane support vector machine[J]. Telecommunications Science, 2015,31(8): 78-83.
[11]	GEORGE A . Anomaly detection based on machine learning dimensionality reduction using PCA and classification using SVM[J]. International Journal of Computer Applications, 2012,47(21): 5-8.
[12]	BAO S , ZHANG L , YANG G . Trajectory outlier detection method based on kernel principal component analysis[J]. Journal of Computer Applications, 2014,34(7): 2107-2110.
[13]	SAKURADA M , YAIRI T . Anomaly detection using autoencoders with nonlinear dimensionality reduction[C]// Mlsda Workshop on Machine Learning for Sensory Data Analysis,December 2,2014,Gold Coast,Australia QLD,Australia. New York:ACM Press, 2014: 4-11.
[14]	HINTON G E . Training products of experts by minimizing contrastive divergence[J]. Neural Computation, 2002,14(8): 1771-1800.
[15]	SUBRAMANIAM S , PALPANAS T , PAPADOPOULOS D ,et al. Online outlier detection in sensor data using non-parametric models[C]// International Conference on Very Large Data Bases,September 12-15,2006,Seoul,Korea. New York:ACM Press, 2006: 187-198.
[16]	MOORE B . Principal component analysis in linear systems:controllability,observability,and model reduction[J]. IEEE Transactions on Automatic Control, 2003,26(1): 17-32.
[17]	HU C , HOU X , LU Y . Improving the architecture of an autoencoder for dimension reduction[C]// Ubiquitous Intelligence and Computing,2014 IEEE,Intl Conf on and IEEE,Intl Conf on and Autonomic and Trusted Computing,and IEEE,Intl Conf on Scalable Computing and Communications and ITS Associated Workshops,Dec 9-12,2014,Bali,Indonesia. Piscataway:IEEE Press, 2014: 855-858.
[18]	HINTON G E . A practical guide to training restricted Boltzmann machines[M]. Berlin: Springer Berlin Heidelberg, 2012: 599-619.
[19]	YANG J , DENG T , SUI R . An adaptive weighted one-class svm for robust outlier detection[M]. Berlin: Springer Berlin Heidelberg, 2016.
[20]	LIN C J . A practical guide to support vector classification[EB/OL]. .

算法模型	数据集				平均值
算法模型	FC	GAS	DSA	HAR	平均值
SVDD	95.45%	91.73%	82.66%	86.73%	89.14%
PCA-SVDD	96.55%	95.23%	91.48%	92.34%	93.90%
DBN-SVDD	97.92%	98.29%	96.76%	97.55%	97.63%

算法模型	数据集				平均值
算法模型	FC	GAS	DSA	HAR	平均值
SVDD	80.34%	83.54%	79.47%	78.62%	80.49%
PCA-SVDD	84.32%	91.45%	79.13%	79.85%	83.69%
DBN-SVDD	98.50%	96.52%	97.14%	98.45%	97.65%

核函数	算法模型	数据集				平均值
核函数	算法模型	FC	GAS	DSA	HAR	平均值
	AE	97.83%	97.54%	96.32%	97.25%	97.24%
RBF	DBN-SVDD	97.92%	98.29%	96.76%	97.55%	97.63%
线性	DBN-SVDD	98.50%	96.52%	97.14%	98.45%	97.65%

核函数	算法模型	数据集				平均值
核函数	算法模型	FC	GAS	DSA	HAR	平均值
线性	DBN-SVDD	0.076 8	0.096 9	0.149 5	0.241 6	0.141 2
RBF	DBN-SVDD	0.090 9	0.107 5	0.188 1	0.318 6	0.176 3
	AE	0.436 8	0.987 4	0.632 5	1.031 3	0.772 1

核函数	算法模型	数据集				平均值
核函数	算法模型	FC	GAS	DSA	HAR	平均值
线性	DBN-SVDD	0.121 3	0.183 1	0.258 1	0.562 8	0.281 3
RBF	DBN-SVDD	0.182 4	0.359 7	0.473 8	0.876 7	0.473 1
	AE	1.537 2	2.243 6	5.569 0	6.622 2	3.993 0

基于深度信念网络和线性单分类SVM的高维异常检测

High-dimensional outlier detection based on deep belief network and linear one-class SVM

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 11

参考文献 20

相关文章 15

Metrics

推荐阅读 0

[1]	陈华华, 陈哲. 基于钉板分布稀疏变分自编码器的异常检测算法研究[J]. 电信科学, 2022, 38(12): 65-77.
[2]	王速, 卢华, 汪硕, 蔡磊, 黄韬. 智能运维中KPI异常检测的研究进展[J]. 电信科学, 2021, 37(5): 42-51.
[3]	陈华华, 陈哲, 郭春生, 应娜, 叶学义, 章坚武. 基于混合高斯变分自编码网络的异常检测算法[J]. 电信科学, 2021, 37(4): 54-61.
[4]	沈潇军,葛亚男,沈志豪,倪阳旦,吕明琪,翁正秋. 一种基于LSTM自动编码机的工业系统异常检测方法[J]. 电信科学, 2020, 36(7): 136-145.
[5]	吴亚萍,董红召,林盈盈,柳菁. 混沌改进鱼群算法及其在工业控制网络异常检测中的应用[J]. 电信科学, 2020, 36(3): 27-33.
[6]	谭晓敏,方艾,金铎,李长江. IPTV用户体验异常的自动化检测[J]. 电信科学, 2019, 35(7): 159-164.
[7]	苏赋,吕沁,罗仁泽. 基于深度学习的图像分类研究综述[J]. 电信科学, 2019, 35(11): 58-74.
[8]	钱途,钱江波,董一鸿,陈华辉. SLSB-forest：高维数据的近似k近邻查询[J]. 电信科学, 2017, 33(9): 58-68.
[9]	彭聪,钱江波,陈华辉,董一鸿. 基于加权自学习散列的高维数据最近邻查询算法[J]. 电信科学, 2017, 33(6): 73-85.
[10]	李斐,陈恳,李萌,郭春梅. 一种基于运动相似熵的人群异常行为检测[J]. 电信科学, 2017, 33(5): 90-98.
[11]	姜红红,张涛,赵新建,钱欣,赵天成,高莉莎. 基于大数据的电力信息网络流量异常检测机制[J]. 电信科学, 2017, 33(3): 134-141.
[12]	李传煌,程成,袁小雍,岑利杰,王伟明. 基于深度学习的软件定义网络应用策略冲突检测方法[J]. 电信科学, 2017, 33(11): 27-36.
[13]	张玉,戴磊. 基于heavy hitter流提取的网络异常检测优化研究[J]. 电信科学, 2010, 26(3): 46-51.
[14]	杨松,李宗林. 一种基于高聚集链路测量的网络级别OD流异常检测框架[J]. 电信科学, 2010, 26(10): 121-126.
[15]	周彬彬,崔宝江,杨义先. 基于系统行为分析的异常检测技术研究[J]. 电信科学, 2009, 25(2): 59-65.