Sei sulla pagina 1di 6

QoS-QoE Correlation Neural Network Modeling for

Mobile Internet Services


Sebastin Rivera, Horacio Riveros, Christian Ariza-Porras, Carlos Lozano-Garzon, Yezid Donoso
Computing and System Engineering Department
Universidad de los Andes, Bogot, Colombia
Email:{s.rivera57, lh.riveros102, cf.ariza975, ca.lozano968, ydonoso}@uniandes.edu.co
of user perception. Nowadays, these operators knows that
user satisfaction is a key success factor for the loyalty and
positioning of the company against its competitors.
This paper shows the validation of the model obtained of
the methodology QoE-QoS decision making tool proposed in
[4]. We are doing a special emphasis in the use of neural
networks as the main tool to correlate the QoS parameters and
the user QoE. The model aim is to support the mobile operator
decisions about the maintenance of network infrastructure, as
well as the expansion of the same, specifically for their MIS.
The remainder of this paper is organized as follows. In
Section 2 we show works related with the measurement of
QoE and the correlation of it with QoS; in Section 3 we present
the QoE-QoS correlation model. Validation of the model and
results are present in Section 4 and finally, Section 5 presents
the conclusions and future works.

AbstractThe users loyalty is a key component to be


considered by the mobile telecommunication operator into
a highly competitive environment with a continuous growth
of mobile internet services. For this reason, we propose a
methodology to support the network performance decision to
provide mobile internet services based in the users perception.
The correlation between the Quality of Service parameters and
the Quality of Experience is the base of this methodology; we
use two different neural network models to obtain the final
correlation model. The model is proved gathering information
of five different cells using a 3G modem and a 3G mobile
phone (Android OS). In this work we are validating the model
obtained only with gathering network performance parameters.
Keywords: Decision making tool, Mobile internet services,
Network performance, Neural networks, Quality of experience,
Quality of service.

I. I NTRODUCTION
The fast evolution of the different technologies involved
in cellular mobile networks has led to the initial provision
of voice service to the possibility of transmitting voice, data
and video over the same network. Nowadays, Mobile Internet
Service (MIS) is one of the most growing services, the International Telecommunication Union (ITU) esteems that there
are about 1.2 billion active mobile-broadband subscriptions
to these services in the world [1]. This growing market has
brought the mobile telecommunications operators to implement network performance optimization mechanisms. These
new mechanisms should considering the ensuring of quality
of service (QoS) and quality of experience (QoE).
The QoS concept was defined by the ITU as "the collective
effect of service performance which determines the satisfaction
of a service user"[2]. Actually, the telecommunication operator
relates this concept only with technical aspects. The QoE concept is closely related with the customer decisions to continue
using a service. This decision depends on the perceived quality,
the price of the service, and supplier responses (problems and
complaints). The guidelines of QoE were defined by the ITU
in [3]. Even though QoS and QoE measurements are quite
different, they have a high degree of correlation; nevertheless,
some mobile operators have not yet implemented tools for
incorporating the feel of a user based on the QoS parameters
measure for a specific service. Formerly, Mobile phone companies did improves in the deployment and delivery of their
products influenced by the quality of service (QoS), regardless

978-1-4673-2088-7/13/$31.00 2013 IEEE

II. R ELATED W ORK


We addressed the review of the related work from two
perspectives. The first, QoS and QoE correlation models
independent of the technique used, and the second the use of
neural networks in problems related to QoS and QoE. Also,
we explore the use of this kind of correlation models as a tool
for the Vertical Handover Decision Algorithm.
A. QoS - QoE correlation works
Some of these works, examine the relation between QoS parameters and user perception for different services. In [13] the
authors propose a model for QoS-QoE correlation reflecting
all parameters that influence QoE in converged networks. The
goal of this work was find the QoS level required to fulfill
the QoE. They assign a weight that involves the key QoS
parameters according to the type of application. Mok, Chan
and Chang [14] conducted a research about the relationship
between the three levels of QoS in HTTP video streaming
(network QoS, application QoS and user QoS) and how
these affect the QoE. The methodology used by the authors
was the correlation between application and network QoS
using analytical models and empirical evaluation. Then they
performed subjective evaluations to correlate QoS and QoE,
the analysis concluded that the frequency of rebuffering is the
main factor responsible for the variations in the QoE.

75

Authors of [3] propose three layers evaluation architecture


in wireless network. The top layer is QoE index which defines
the characteristics of sensations, perceptions and opinions of
users who are interacting. The middle layer is QoE parameters
and describes the information related to customer experience
obtained through surveys. Customer satisfaction can be predicted exactly by the analysis of that information. The third
layer is QoS parameters that reflect the service capability
and are generally referred to quality of end-to-end service.
In this work the evaluation algorithm based on fuzzy analytic
hierarchy process is discussed and applied on the WAP service.
The results show that with QoS parameters are correlated with
greater weight to the perception of the user.
Wei, Xu, Wen, Liu and Yan in [15] develop a new subjective
assessment method User Satisfaction Threshold Measurement
(USTM). They also develop an online test agent on smart
phones to let the users join the test at any places at any
time, in the framework for subjective assessment distorted
video samples are generated by mapping from a simple Endto-End QoS metric set. As a result of this assessment four
user satisfaction thresholds, multi-level user opinion, and their
mapping to the parameters in End-to-End QoS are obtained.

well as the network parameters affecting the video quality and


QoE.
As a final applications, in [11] Piamrat, Ksentini and
Bonnin propose an admission control mechanism based on
QoE perceived by users to connect into a WLAN Access
Point, to guarantee service quality at users and to optimize
resource utilization. The QoE is obtained by PSQA. The
authors of [12] propose the use of a multilayer artificial neural
network to doing an estimation of Quality of Experience (QoE)
metrics based on Quality of Service (QoS) metrics in WiMAX
networks; from the results obtained, it was observed that the
neural network for the scenario in question had a very good
prediction.
C. QoS-QoE correlation model as the base of Vertical Handover Decision Algorithms.
As was identified in [23], one of the most challenging
problems for the mobile telecommunication operators is maintaining the service continuity across the new network infrastructure. This infrastructure must provide Vertical Handover,
that is understood as the possibility that one user equipment
can change their access network without loss of the service. A
lot of works have been developed around this topic, but only a
few of them propose the use of QoE as a decision parameter.
The use of Quality of Experience as a decision parameter its
quite used in researchs around the transmission of multimedia
services, specially in IEEE network environments (802.11 and
802.16). In [24] the authors proposed a network selection
mechanism based QoE and the traffic load of each network.
This study shown a gain of 12.5% in the video perceived quality. Cerqueira et al. in [25] presents a Quality of Experience
Handover Architecture for Converged Heterogeneous Wireless
Networks. This proposal is an extension to IEEE 802.21 Media
Independent Handover, to maximize the user QoE.
Authors of [26] proposed a user-based network selection
mechanism. They modeled this mechanism as a multi criteria
problem and they used TOPSIS to solve it. Also its important to mention is that QoE parameter is obtained through
the implementation of Pseudo-Subjective Quality Assessment
(PSQA) technique. The preliminary results shown that this
QoE-aware mechanism improve significantly the user experience and maintain the load balancing between networks.
Zekri, Pokhrel, Jouaber and Zeghlache in [27] proposed a
reputation based Vertical Handover Decision mechanism. This
solution combining the use of reputation as a QoE indicator
and the Stream Control Transmission Protocol (SCTP) as a
mobility protocol. Performance results show that the proposed
solution allow seamless mobility with lowest delays and better
throughput.
The related work show the important issue to generate a
QoE-QoS correlation model over wired and wireless networks,
and how the use of Neural Networks are a key tool in this task,
based on its high performance and low cost. Our proposal
aims to establish a QoE-QoS correlation model for mobile
internet services (Web Browsing, Video Streaming and File
Downloads) by using the ANN.

B. Use of Neural Network related with QoS and/or QoE


One of the most common applications of the Neural Networks related with QoS and/or QoE is the selection of the best
available web service. In [5] Guo, Chen, Yang and Fei propose
a principal component analysis and neural network algorithm
to adjust QoS attributes weight dynamically, objectively and
automatically in order to find the best available web service.
Li and Yan-xiang in [6] propose an independent component
analysis and artificial neural networks method to build up an
automatic and objective web service ranking model. Finally
in [7] the authors shown a wavelet neural network (WNN)
application to predict the QoS of Web Service in order to
choose the best.
Other papers examine the correlation between QoS and QoE
specifically in Multimedia Services. In [9] the authors propose
a Back-Propagation Neural Network Algorithm to connect
QoE directly to QoS according to the corresponding level
of QoE under different network circumstance; they test their
proposal with the help of Video Quality Evaluation System and
Network Emulator. Rubino, Tirilly and Varela [8] proposed
a methodology called Pseudo-Subjective Quality Assessment
(PSQA), based on Random Neural Networks, which is able to
provide an accurate and automatic quantitative evaluation of
the perceived quality of an audio or video communication over
a packet network, where the flow is subject to different kinds
of distortions. Singh, Ksentini, and Marienval [9] designed an
automatic QoE measurement tool for Scalable Video Coding
(SVC) video coding mechanism. The proposed module is
based on PSQA. In [10] Cherif, Ksentini, Negru, and Sidibe,
present a new QoE tool solution, named ALICANTE Pseudo
Subjective Quality Assessment (A_PSQA). It relies on a NoReference QoE measuring approach, fully functional in Future
Media Internet context relation between the video coding as

76

III. Q O S - Q O E C ORRELATION MODEL CONSTRUCTION

A. Multilayer Perceptron
The Multilayer Perceptron (MLP) neural network is made
up of multiple layers of nodes in a directed graph and each
layer is connected to the next, which can solve problems that
are not linearly separable.
The MLP uses a supervised learning technique called
backpropagation for the training of the network [19], [20].
The backpropagation algorithm is used for the process of
pattern recognition supervised learning and interpolation, also
is useful in research to solve problems stochastically.
When observing the multilayer perceptron architecture evidence that multiple inputs are mapped based on the outputs
of the intermediate layers and parameters. The following
properties were shown in the construction architecture of a
multilayer perceptron:
No connections within a layer
No direct connections between input and output layers
Fully connected between layers
Often more than 3 layers
Number of output units need not equal number of input
units
Number of hidden units per layer can be more or less
than input or output units

In [4] we propose the methodology show in the Figure 1.


The goal of this development is create a QoS-QoE correlation
model to support the network performance making decision
for critical web services. As relevant aspect of the proposed
methodology is important to emphasize the development of the
subjective test for the evaluation of web services and the use
of neural networks to construct the correlation model between
QoS and QoE.

B. RBFNetwork

Fig. 1.

The RBFNetwork is an ANN, based on radial basis functions(rbf), whose values depends on the distance from the
origin or on the distance from some other point c, (the
norm is usually Euclidean distance). The RBFNetwork is
embedded in a two layer neural network, where each hidden
unit implements a rbf. The output unit implements a weighted
sum of hidden unit outputs. The input is nonlinear, while
the output is linear, due to their nonlinear approximation
properties, this network is able to model complex mappings,
which perceptron neural networks can only model by means
of multiple intermediary layers [21].
The process of network training is done by finding the RBF
weights [22]. If we have a training set, the network parameters
are optimized in order to fit the network outputs to the given
inputs. After this training, the RBF network can be used in online training to adapt the network parameters to the changing
data.

Methodology proposed in [4]

According to [17], the inputs selected for our model are:


bandwidth, latency, and signal strength. These parameters were
obtained by a mobile agent developed for that purpose. The
measurements were sent to a data server using a Web Service
through the mobile cellular network. Once we collected a several number of samples, data were passed through a statistical
process in order to detect and remove those which present data
out of range (three times the standard deviation). The data that
passed this criterion was used as the input parameters for our
neural network. The results obtained will correspond to one
value of QoE (excellent, very good, good, fair or poor).
To select the best classifier for the dataset we did a Weka
experiments. In these experiments we compared different
classifiers, not only based on neural networks, also some
clusters based classifiers. We select the Multilayer Perceptron
and the Radial Basis Function Network (RBFNetwork) as
neural network based classifiers .

C. QoS - QoE correlation model


In order to training the neural network, we decide to use
an approach of living labs. We select five differents points
of the city that represent each MOS values (excelent, very
good, good, fair and poor) according only to QoS. These
selected point was previously characterized by one real Mobile
Telecommunication Operator.We developed a survey to gather
QoE data from people using mobile devices with a special
mobile agent developed. The people browse with mobile
smartphones in four different web contexts (text-only page, a
page with images and text, a video and file download) through

77

TABLE V
C ONFUSION M ATRIX , RBFN ETWORK , ANDROID DATA

an Android Smartphone or a laptop connected to an USB


modem.
You can see some statistics values for the modem gathered
data in Table I. Both neural network classifiers reach good
training results. The modem models were trained using 84
instances of which 71 (84.5238%) was well rated by both
Multilayer percetron and RBFNetwork classifiers, although
they have different confusion matrices as shown in Table II
and Table III.
The letters a, b, c, d, e in Tables II, III and V, represents the
the MOS values. In these matrices you can see how many of
the training data was classified for each class. If exists more
than one value for the same row, the classifier chooses the
class with greater number of data training data.

a
28
0
11
4
0

b
0
2
0
0
0

c
11
0
31
4
0

d
0
1
1
15
0

e
0
0
0
0
2

-classified as
a = 4.5
b=5
c=4
d = 3.5
e = 3.0

data traffic in the cells are not constant in time. Exist some
peak hours when there are more users in the cell and these
use their mobile devices most frequently. The QoE values were
given by different users following the methodology described
in [4].

TABLE I
M ODEM Q O S DATA CORRELATION WITH Q O E

AVERAGE
STDEV
Correlation with QoE

Bandwith
188.23
88.90
0.32

Latency
255.57
365.94
-0.39

Signal Strength
22.37
4.93
-0.003

QoE
3.72
0.5
1

TABLE II
C ONFUSION M ATRIX , M ULTILAYER P ERCEPTRON , MODEM DATA
a
3
0
0
0
0

b
0
12
0
0
0

c
0
0
17
1
0

d
0
0
5
33
4

e
0
0
1
2
6

-classified as
a = 2.5
b=3
c = 3.5
d=4
e = 4.5

Fig. 2.

Bandwith (kbps) vs. QoE (MOS), Android Training Data

TABLE III
C ONFUSION M ATRIX , RBFN ETWORK , MODEM DATA
a
3
0
0
0
0

b
0
7
0
0
0

c
0
1
22
2
0

d
0
4
1
34
5

e
0
0
0
0
5

-classified as
a = 2.5
b=3
c = 3.5
d=4
e = 4.5

As you can see in Table IV, only the RBFNetwork classifier


obtained an acceptable result. The android model was trained
using 113 instances of which 78 (69.0265 %) was well rated
by the classifier, the confusion matrix is shown in Table V.
TABLE IV
A NDROID Q O S DATA CORRELATION WITH Q O E

AVERAGE
STDEV
Correlation with QoE

Bandwith
5080.5
2067.53
-0.01

Latency
261.61
98.8
0.03

Signal Strength
19.20
3.0
-0.006

Fig. 3.

QoE
4.06
0.44
1

Delay (ms) vs. QoE (MOS), Android Training Data

IV. VALIDATION S CENARIO AND R ESULTS


After the validation of the correlation model obtained
through simulations, we propose a living lab environment to
validate it. In this environment we prove the model in two real
operator mobile networks in the same five points defined in the
training phase. We gathered network performance data using
two different android devices. These data was took along all
day just to guarantee an observation of the daily performance
of the network. A total of four hundred samples were taken,

The results of training could be seen in Figure 2 and Figure


3. In both boxplots for a single QoE value, exist a high
variation in the data gathered for bandwith and delay. The
reason for this variation is related with the data measurements.
These were taken all the day without take into account that

78

this number was obtained after an statistical analysis to ensure


a confidential interval of at least 95%. You can see the
distribution of the data obtained in Figure 4 and Figure 5.

relevant in user perception than the bandwith, and the user


perception can not be assessed from a single parameter of
network performance.

Fig. 7. Average Delay (ms) vs. QoE (MOS), Android Trained and Validation
Data
Fig. 4.

Bandwith (kbps) vs. QoE (MOS), Android Validation Data

The experiments show that exist differences between the


average trained data and the average validation data. In order
to know how big was the difference between them and define
if our model was well trained, we used the relative error
metric. The values can be seen in Table VI, and in Figure
8. As you can see for bandwith and delay values, the relative
error decrease for QoE values of 4, 4.5 and 5, the reason for
this could be explained for the number of the trained data.
In the Table V we can see that the most of the users QoE
classification are in this values.

Fig. 5.

Delay (ms) vs. QoE (MOS), Android Validation data

The behaviour of the validation data was as we expect, a


direct relationship between the available bandwidth and QoE,
and an inverse relationship between delay and QoE for MIS,
as seen in Figure 6 and Figure 7.
Fig. 8.

Relative Error, Android data

TABLE VI
R ELATIVE E RROR BETWEEN TRAINING DATA AND TEST DATA FOR EACH
Q O E VALUE
QoE
3.5
4
4.5
5

Bandwith
-0.567
-0.36
-0.421
-0.132

Delay
0.82
0.156
0.125
0.078

Fig. 6. Average Bandwith (kbps) vs. QoE (MOS), Android Training and
Validation Data

V. C ONCLUSIONS

The average training data graphic for bandwith shows an


untypical behavior, it was expected that for a better bandwith
value, the user perception should be better. However the
graphic does not show that, but the validation data shows
correctly that direct relationship. We need to review carefully
the classification model for bandwith and QoE, but in fact we
suggest that for web browsing, the delay parameter is more

This paper proposes the use of neural networks for the


construction of the correlation model between QoS and QoE
for mobile internet services. This model is the base for building
a making decision tool, with which the mobile telecommunication operators can support their decisions about the
optimization of their network performance. In fact, the use of

79

neural networks allow operators to gather data constantly and


prevent rejections by the users in their products or services.
In particular, we focused on Web Browsing service, but our
methodology could in principle be applied to other mobile applications or services. We have tested our methodology in real
life scenarios, throught a real telecommunication operators
network. After looking the training data, and the validation
data, the relationships between networks parameters (bandwith
and delay) and QoE, was as expected, a direct relationship
between bandwith and QoE, and a inverse relationship between
delay and QoE. It is also clear to us that the user perception
for web browsing is not only assessed by these two network
parameters, however this initial study led us to construct a
better correlation model, using neural networks, which show
a good behavior and results in this research.
In summary, considering the use of neural networks to
reduce the gap between the users perception of the service
and the operators offered service is an important improvement
in the telecommunication industry where the fast and constant
evolution in mobile devices and access networks, has led the
sector to a converged environment that enables businesses to
provide new services.

[8] G. Rubino, P. Tirilly, and M. Varela, Evaluating Users Satisfaction


in Packet Networks Using Random Neural Networks, in International
Conference on Artificial Neural Networks, Athens, Greece, 2006, pp.
303 - 312.
[9] K.D Singh, A. Ksentini, and B. Marienval, Quality of Experience
Measurement Tool for SVC Video Coding, in 2011 IEEE International
Conference on Communications (ICC), Kyoto, Japan, 2011, pp. 1 - 5.
[10] W Cherif, A Ksentini, D Negru, and M. Sidibe, A_PSQA: EFFICIENT
REAL-TIME VIDEO STREAMING QOE TOOL IN A FUTURE MEDIA,
in 2011 IEEE International Conference on Multimedia and Expo (ICME),
Barcelona, Spain , 2011, pp. 1 - 6.
[11] K. Piamrat, A. Ksentini, C. Viho, and J.-M. Bonnin, QoE-Aware
Admission Control for Multimedia Applications in IEEE 802.11 Wireless
Networks, in IEEE 68th Vehicular Technology Conference, 2008. , 2008,
pp. 1 - 5.
[12] V.A. Machado et al., A new proposal to provide estimation of QoS
and QoE over WiMAX networks: An approach based on computational intelligence and discrete-event simulation, in 2011 IEEE LatinAmerican Conference on Communications (LATINCOM), Belm do
Par, Brazil, 2011, pp. 1 - 6.
[13] J. K, Hyun, H. L. Dong, M. L.: Kyoung, H. L. Jong, L Won, and Seong
G. C., The QoE Evaluation Method through the QoS-QoE Correlation
Model, in Fourth International Conference on Networked Computing and
Advanced Information Management, Gyeongju, Korea, 2008, pp. 719 725.
[14] M Ricky, C Edmond, and C. Rocky, Measuring the Quality of Experience of HTTP Video Streaming, in Integrated Network Management (IM),
2011, pp. 1 - 8.
[15] K Hornik, M Stinchcombe, and H. White, Multilayer Feedforward
Networks are Universal Approximators, Neural Networks, vol. 2, pp. 359
- 366, 1989.
[16] D.S. Broomhead and D. Lowe, Multivariable functional interpolation
and adaptive networks, Complex Systems, vol. 2, pp. 321 - 355, 1988.
[17] European Telecommunications Standards Institute, Speech Processing,
Transmission and Quality Aspects (STQ); User related QoS parameter
definitions and measurements., European Telecommunications Standards
Institute, Sophia Antipolis Cedex - FRANCE, Standard ETSI EG 202
057-2 V1.3.1, 2009.
[18] H. Du, C. Guo, Y. Liu, and Y. Liu, Research on relationship between
QoE and QoS based on BP Neural Network, in IEEE International
Conference on Network Infrastructure and Digital Content, 2009. ICNIDC 2009. , Bejing, China, 209, pp. 312 - 315.
[19] Rosenblatt, Frank. x. Principles of Neurodynamics: Perceptrons and the
Theory of Brain Mechanisms., Spartan Books, Washington DC, 1961.
[20] Rumelhart, David E., Geoffrey E. Hinton, and R. J. Williams. Learning
Internal Representations by Error Propagation. David E. Rumelhart,
James L. McClelland, and the PDP research group. (editors), Parallel
distributed processing: Explorations in the microstructure of cognition,
Volume 1: Foundations. MIT Press, 1986.
[21] Park,J.,Sandberg, J. W.,Universal approximation using radial basis
functions network, Neural Computation, vol.3,pp. 246-257, 1991.
[22] Adrian G. Bors, Introduction to the Radial Basis Function (RBF)
Networks, OnLine Symposium for Electronics Engineers. 2007.
[23] Lozano C, Ortiz N, Donoso Y., A Proactive Vertical Handover Decision
Algorithm in Heterogeneous Wireless Networks for Critical Services,
International Journal of Computers, Communications & Control , vol.
VIII, no. 3, 2013.
[24] Valenzuela G, Ferreira I, Nobrega O, Cunha P., Vertical Handover
Decision Based on Quality of Experience in Heterogeneous Wireless
Networks, Proceedings of the 6th Euro American Conference on Telematics and Information Systems (EATIS 12), Rogerio Patricio Chagas do
Nascimento (Ed.). ACM, New York, NY, USA, 2012.
[25] Cerqueira E, Quadros C, Venancio-Neto A, Riker A, Immich R,
Curado M, Pescapl A., Quality of Experience Handover System for
Heterogeneous Multimedia Wireless Networks, International Conference
on Computing, Networking and Communications 2013 (ICNC), Wireless
Networks Applications Track, San Diego, CA, USA, January 2013.
[26] Piamrat K, Ksentini A, Viho C, Bonnin J.-M., QoE-aware vertical
handover in wireless heterogeneous networks, 7th International Wireless
Communications and Mobile Computing Conference (IWCMC), 2011.
[27] Zekri M, Pokhrel J, Jouaber B, Zeghlache D., Reputation for Vertical
Handover decision making, 17th Asia-Pacific Conference on Communications (APCC), 2011.

VI. F UTURE W ORK


In the study, we detect the need to considering the data
traffic variability in the network along the day in the correlation model. Therefore, we are currently extending this study
using time series in order to understand better this phenomen
and its impact in the model. Also, it is necessary validate the
methodology and the model developed under the new network
access platforms such as the Long Term Evolution (LTE). After
that, in order to improve the significant impact of this research
in the mobile network operators, our next step is the develop
of two specific making decision tools one for O&M and other
for vertical handover decision.
R EFERENCES
[1] International Telecommunication Union , The World in 2011 - ICT Facts
and Figures, Geneva, Switzerland, 2011.
[2] Telecommunication Standardization Sector (ITU-T), Terms and Definitions Related to Quality of Service and Network Performance including
Dependability, International Telecommunication Union, Geneva, Switzerland, Recommendation E.800, 1995.
[3] Telecommunication Standardization Sector (ITU-T), Methods for subjective determination of transmission quality, International Telecommunication Union, Geneva, Switzerland, Recommendation P.800, 1996.
[4] C Lozano-Garzon, C Ariza-Porras, S Rivera-Diaz, H Riveros-Ardila,
and Y. Donoso, Mobile Network QoE-QoS Decision Making Tool for
Performance Optimization in Critical Web Service, International Journal
of Computers, Communications & Control , vol. VII, no. 5, November
2012.
[5] L. Guo, H. Chen, G. Yang, and R. Fei, A QoS Evaluation Algorithm
for Web Service Ranking Based on Artificial Neural Network, in 2008
International Conference on Computer Science and Software Engineering,
Wuhan, China, 2008, pp. 381-384.
[6] W. Li and H. Yan-xiang, A New Web Service Ranking Method Based on
QoS Independent Components and ANN, in 2010 6th International Conference on Wireless Communications Networking and Mobile Computing
(WiCOM), Chengdu, China, 2010, pp. 1 - 4.
[7] Z. Jin-hong, A Short-Term Prediction for QoS of Web Service Based on
Wavelet Neural Networks, in 2010 International Conference on Computational Intelligence and Communication Networks (CICN), Bhopal, India,
2010, pp. 254 - 258.

80

Potrebbero piacerti anche