第
46
卷第
2
期
2012
年
2
月
西安交通大学学报
JOURNAL
OF
XI'AN
JIAOTONG
UNIVERSITY
采用两阶段策略模型
(KTSVM)
的
P2P
流量识别方法
丁要军1,
2
蔡皖东
l
Vo
l.
46
No.2
Feb.
2012
(1.西北工业大学计算机学院,
710129
,西安;
2.
咸阳师范学院信息工程学院,
712000
,陕西咸阳)
摘要:针对识别加密
P2P
网络流量比较困难的问题,提出一种基于
K
均值和直推式支持向量机
CTSVM)
的半监督学习模型
两阶段策略模型
CKTSVM
,
k-means
based
transductive
support
vector
machine)
,以提高
P2P
流量的识别精度.该模型首先使用
K
均值半监督聚类算法计算训练
集中正例样本的数目,然后根据正例样本的数目来训练、
TSVM
分类模型,提高了
TSVM
模型的稳
定性和准确性.该模型的优势是可以使用未标注样本和标注样本共同训练分类模型,非常适合于识
别标注比较困难的
P2P
流量.实验结果表明,在标注样本较少的情况下,该模型的识别精度和稳定
性均优于
TSVM
模型和
SVM
模型.
关键词:直推式支持向量机;半监督学习;流量识别;对等网络流量;王联网
中图分类号
TP393
文献标志码
A
文章编号:
0253-987X(2012)02-0045-06
P2P
Traffic Identification via k-Means ßased
Transductive
Support
飞
T
etor Machine
DING
Yaoj
un
l,
2
,
CAI
Wandong
l
O.
School
of
Co
mputer
Science
and
Technology
,
Northwestern
Polytechnical
University
,
Xi'
an
710129
,
China;
2.
School
of
Information
Engineering
,
Xianyang
Normal
University
,
Xianyang
,
Shaanxi
712000
,
China)
Abstract:
A
new
semi-supervised
learning
model
based
on
k-means
and
transductive
support
vec
tor
machine
is
proposed
to
improve
the
accuracy of
P2P
traffic identification.
The
semi-supervised
cluster
algorithm
of
k-means
is
used
to
calculate
the
number
of
positive
instances
in a
training
set
,
and
then
the
TSVM
model
is
trained
based
on
the
number
of
positive instances.
So
the
stability
and
accuracy
of
TSVM
are
improved.
An
important
advantage
of
the
model is
that
the
model
can
be
trained
by
both
labeled
samples
and
unlabeled
samples
,
and
the
model
is
suitable
for
the
iden
tification of
P2P
traffic
that
is difficult
to
be
labeled.
Experimental
results
show
that
the
proposed
model
is
better
than
TSVM
and
SVM
models
in accuracy
and
stability
,
and
that
it
is
an
effective
way
to
improve
the
accuracy
of
P2P
traffic
identification.
Keywords:
transductive
support
vector
machine;
semi-supervised
learning;
traffic
identification;
peer-to-peer
traffic;
In
ternet
随着
P2PC
peer-to-peer)
网络技术的广泛应用,
P2P
已经取代
www
应用成为占用带宽最多的应
用协议,
P2P
流量在中国教育科研网的骨干网上所
占的比例已经从过去的
0.76%
激增到
70%
左右
[IJ
收稿日期:
2011-06-30.
作者简介
2
丁要军
0980-)
,男,博士生;蔡皖东(通信作者)
,男,教授,博士生导师.
基金项目:国
家高技术研究发展计划资助项目
(2009AA01Z424)
;西北工业大学基础研究基金资助项目(J
C201149)
;咸阳师范学院专项科
研基金资助项目
(07XSYK268).
网络出版时间:
2011-12-01
网络出版地址:
http://www.cnki.net/kcms/detai
I/
6
1.
1069.
T.
2011120
1.
1640.001.
html
评论0
最新资源