An effective method for clustering-based web service recommendation

International Journal of Electrical and Computer Engineering (IJECE)
Vol. 12, No. 2, April 2022, pp. 1571~1578
ISSN: 2088-8708, DOI: 10.11591/ijece.v12i2.pp1571-1578  1571
Journal homepage: http://ijece.iaescore.com
An effective method for clustering-based web service
recommendation
Ha Huy Cuong Nguyen1
, Bui Thanh Khiet2
, Van Loi Nguyen1
, Thanh Thuy Nguyen3
1
Faculty of Computer Science, Vietnam - Korea University of Information and Communication Technology, The University of Danang,
Danang, Vietnam
2
Institute of Engineering and Technology, Thu Dau Mot University, Binhduong, Vietnam.
3
Faculty of Computer Science, VNU University of Engineering and Technology, Hanoi, Vietnam
Article Info ABSTRACT
Article history:
Received Jun 28, 2021
Revised Sep 16, 2021
Accepted Oct 10, 2021
Normally web services are classified by the quality of services; however, the
term quality is not absolute and defined relatively. The quality of web services
is measured or derived using various parameters like reliability, scalability,
flexibility, and availability. The limitation of the methods employing these
parameters is that sometimes they are producing similar web services in
recommendation lists. To address this research problem, the novel improved
clustering-based web service recommendation method is proposed in this
paper. This approach is mainly dealing with producing diversity in the results
of web service recommendations. In this method, functional interest, quality
of service (QoS) preference, and diversity features are combined to produce a
unique recommendation list of web services to end-users. To produce the
unique recommendation results, we propose a varied web service
classification order that is clustering-based on web services’ functional
relevance such as non-useful pertinence, recorded client intrigue importance,
and potential client intrigue significance. Additionally, to further improve the
performance of this approach, we designed web service graph construction,
an algorithm of various widths clustering. This approach serves to enhance
the exceptional quality, that is, the accuracy of web service recommendation
outcomes. The performance of this method was implemented and evaluated
against existing systems for precision, and f-score performance metrics, using
the research datasets.
Keywords:
QoS prediction
Service-oriented computing
performance
Web service recommendation
clustering
This is an open access article under the CC BY-SA license.
Corresponding Author:
Bui Thanh Khiet
Institute of Engineering and Technology, Thu Dau Mot University
Binhduong, Vietnam
Email: khietbt@tdmu.edu.vn
1. INTRODUCTION
To expand the host pool, the client sends a solicitation and acquires a reaction from utilizing web
hosting. Fundamentally, web hosting can be adopted in two distinct ways. They can be utilized as
straightforward web hosting that give an interface to get information sources and return yields or they can be
utilized as segments that can be incorporated into business forms. The first type of usage is called individual
use and the second is referred to as process use. This research work deals with recommending services with
respect to the individual case.
Though web service technologies and service-oriented computing (SOC) promise to provide loose
coupling among parts and dexterity to react to changes in necessities with obviously conveyed registering and
lesser progressing ventures, wavelength routed (WS) is not shared and reused as expected [1]. One of the

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 12, No. 2, April 2022: 1571-1578
1572
reasons that impede the usage of such technologies and SOC is that efficient WS discovery presents many
challenges [2]-[5]. Recommender systems (RS) are one tool to help bridge this gap. There are various
mechanisms being employed to create RS and the common systems include two main classes, namely, content
basis and collaborative filtering schemes. Content basis RS does the matching between textual information of
a particular product with the textual information representing the interests of a customer. Collaborative filtering
methods use designs in customer grades to recommend. Both types of RS expect notable data resources under
the order of user ranks and product features; hence they are not able to generate high-quality recommendations.
Web service is related to a web services description language (WSDL) description that includes the
depiction of the service. Various studies have been performed to use WSDL archives [6]-[9]. Kumar et al. [4]
suggested that WS index Google is reasonable for giving WS closeness search. In a few cases, their gadget
does not adequately acknowledge data types, which for the most part uncover significant data regarding the
operation of web services [8]. Elgazzar et al. [9], Liu et al. [10] exhibited a relative technique, which conveys
WSDL reports to build the non-semantic website composition appearing. They collected singular segments in
WSDL records as their segments and organized web services into worth-based social issues. The clustering
impacts compartment was utilized to build up the idea of web service record data.
Lausen and Haselwanter [11] executed substance mining systems to concentrate highlights, for
example, service content, setting, hostname, and name, from web service portrayal documents so as to bunch
web services. They proposed a consolidated component burrowing and the clustering approach concerning web
services as a predecessor to exposure in order to assist in constructing a web search contraption to the edge and
push non-semantic web services [12]-[15]. Maximilien and Singh [16], Sofian et al. [17] proposed a multi-
master based structure where managers help quality-based organization certification using an office to disperse
reputation and support data. Every go-between service is independent, yet furthermore cooperates with various
pros to accumulate diverse suppositions and along these lines intensifies its data to improve its fundamental
pro. Lausen and Haselwanter [11] executed the calculation about how to solidify undeniable quality of service
(QoS) estimations to get a sense as a rule rating for a web service. The proposed reputation can be portrayed
as the standard given to the organization by the end client [18]-[20].
This work proposes methods for service discovery that are lighter than those based on semantics and
can be a feasible way towards the realization of service-oriented applications. We attempt to overcome the
difficulties of forecasting QoS values by combining Pearson similarity and the slope one method. Our simple
enhanced algorithm for ranking services considering users’ requirements is better than the existing complicated
algorithms. The basic purposes of this research are: i) to propose an approach to build a semantic kernel
consisting of semantically similar web hosting using the various widths clustering and merging method; ii) to
design new efficient and scalable algorithms for various widths clustering based web service reliability for the
recommendation systems; and iii) to design, implement, and evaluate the proposed technique. This paper is
organized as follows: section 1 presents the introduction and the related work. Section 2 introduces the
methodology. Section 3 shows results and discussions. Finally, conclusions are presented in section 4.
2. PROPOSED METHOD
We are contributing to the existing web service recommendation approaches with the proposed
algorithm called clustering-based system to overcome the limitation of web service recommendation. The
advanced strategy will be used to develop the production of the system. This system is shown in Figure 1. All
the functionalities for web services on devices (WSD) are used by the proposed method CWBR with one extra
functionality, that is, clustered data. Below we define all functionalities of the method.
2.1. Functional evaluation
The functional appraisal can be furthermore isolated into two sections; functional estimation and
utilitarian estimation. Functional estimation considers the result of the client’s chronicled expectation with web
services controlled to a premise-based equivalence criterion. The substance-based identity is procured by object
closeness. This process simply recognizes web services that are represented by the WSD. All things considered;
it is anything but difficult to stretch out our work to deal with different sorts of Web services. The client’s real
interest can be mined from his/her very own affiliation use or requesting history. Utilitarian estimation predicts
the client’s potential interest and diagrams its congruity with web services by using shared isolating based on
customer comparability. This comparability is estimated depending on the web hosting summon history of all
web hosting clients.
2.2. Non-functional evaluation
Think about that m QoS structures are working for estimating the non-utilitarian quality of USi, its
QoS vector is meant by SWi, i.e., SWi=(𝑞𝑖, 1 ,𝑞𝑖, 2, …, 𝑞𝑖, 𝑚), where 𝑞𝑖, 𝑗 expresses the value of the ij quality

Int J Elec & Comp Eng ISSN: 2088-8708 
An effective method for clustering-based web service recommendation (Ha Huy Cuong Nguyen)
1573
standard. For the most part, there are two types of QoS measures. A QoS model is seen as negative if the bigger
the worth, the lower the quality (e.g., cost besides reaction time). Otherwise, the QoS measure is seen as
positive (e.g., Accessibility and Unwavering quality). Evaluations of different QoS criteria should be built up
to a tantamount arrangement for different assessment purposes. While having a previous uniformity, it is
feasible to implement the measurable strategy (i.e., Pauta Paradigm methodology) before the procedure of QoS
esteems ahead of time to expel the exceptions. Here, to change each QoS standard incentive to a genuine
number somewhere in the range of 0 and 1, contrasting it and the base and most extreme estimations of the
QoS basis among all accessible web hosting up-and-comers are needed. After such standardization preparation,
the more noteworthy incentive for the quality is, the more a model implies excellent quality.
Figure 1. System architecture
Step 1. Input dataset: i) user set, ii) web service set, and iii) QoS matrix
Equation (1) userSimda,db=2×|USab|/(|Sda|+|Sdb|), where Sda and Sdb are the numbers of web
services appropriated by the user da and db respectively, USab denotes the collection of Web services utilized
with both da and db, i.e., USab=Sda ∩Sdb. If USab=0, then use (ua,ub)=0.
Equation (2) (US𝑖, US𝑗)=𝜑𝑡𝑒𝑥𝑆𝑖𝑚 +𝜙𝑜𝑝𝑆𝑖𝑚, where 𝑡𝑒𝑥𝑆𝑖𝑚=cos(𝒘𝑖,wj)=𝒘𝑖∙𝒘𝑗/|𝒘𝑖|×|𝒘𝑗| where
|𝒘𝑖| and |𝒘𝑗| signify the euclidean length of the vector 𝒘𝑖 and wj respectively. Moreover, the numerator is the
dot outcome of 𝒘𝑖 with 𝒘𝑗. Where, opSim=
Step 2. Various widths clustering
In this subsection, the various widths clustering part is explained, where an instructive file is
appropriated into a few packages whose size is required by the client characterized limit. Three procedures are
engaged with this activity: getting the group width, dividing, and blending. They are executed successively
until the criteria are met. That is, the procedures of isolating and blending are ceased when the size of the
greatest gathering isn't actually a customer described edge b, or when the quantity of packs with the
accomplishment of the referenced methods is comparable to b. Algorithm 2 demonstrates the midpoints of the
referenced plans, with the factors, information structures, and capacities utilized by the calculation.
Step 3. Cluster-width learning
Given D, an information collection to be clustered, including NNk(Hi) be the method of k-nearest
neighbors for the target Hi, and clsWidth be the method calculating the width (radius) of NNk(Hi), to find the
appropriate worldwide width, two or three entities from D,H={H1;H2; ...;Hr) where r<|D| are randomly chosen,
and for each entity, the span of its k-nearest neighbors is enlisted, and the ordinary is used as an overall width
for D as sought after:

 ISSN: 2088-8708
1574
𝑤 =
1
𝑟
∑ 𝑐𝑙𝑠𝑊𝑖𝑑𝑡ℎ(𝑁𝑁𝐾(𝐻𝑖)))
𝑟
𝑖=1 (1)
This procedure segments a dataset into various clusters utilizing a huge width to arrange the result of
clustering the meagerly distributed items in the n-dimensional range. Be that as it may, large clusters from
thick territories will be made, for example, clusters C2 and C3. Along these lines, every huge bunch whose
size surpasses a client characterized limit (greatest group size) will be separated into various clusters utilizing
a width that changes the depth of that collection. This method progresses continuously until all clusters are not
greater than the client characterized limit. The delivered clusters utilize various-widths clustering web hosting
suggestion framework, where large clusters are parceled into various smaller clusters. The primary steps of this
technique are condensed in the framework for coursing in algorithm 1. This method has two elements:
Gathering additionally b. The past is a look at class objects, where everything contains characters similar
properties of a gathering. In the fundamental development, the entire informative record is perceived as a
gathering and it’s driven whose width is fixed with zeros (Stage 4). The last factor is the division to the best
packs size. At the next stage, the capacity biggest group restores the biggest bunch U, which isn’t allocated as
non-divisional, from clusters (Stage 14). On the off chance that the size regarding U is more remarkable than
(or grows to) b, State (1) implies appropriation to figure a suitable width w for apportioning U. In the event
when the estimation of w is equal to zero, U is assigned as non-disseminated (Stages 15-20). This occurs when
the items in U have similarities regarding the separation work, and in this manner, they can’t be apportioned.
Otherwise, Algorithm 1 is charged into division U (Stage 21). On the off chance that the quantity of delivered
clusters is only one, the estimation of w is huge and it ought to be limited by 10% and utilized once more
(Stage 27). Otherwise, the new groups conveyed from U are added to bunches as opposed to U, and the greatest
pack again is pulled from bunches (Stages 22-25). The processes (Stages 15-27) are repeated until the partition
of the biggest bundle in Groups is less than b.
2.3. Algorithms
Algorithm: 1 Various-widths clustering
Input: Data; US𝑢,1, US𝑢,2, ⋯, US𝑢,𝑀; 𝑷𝑢,1,𝑷𝑢,2,⋯,𝑷𝑢,𝑀; 𝜀; US1,US2,⋯,US𝑁; 𝑆𝑊1,𝑆𝑊2,…..,SW𝑁; α
Output: 𝐶𝑙𝑢𝑠𝑡𝑒𝑟𝑠; 𝑈𝑢,1, 𝑈𝑢,2, ⋯, 𝑈𝑢,𝑁
1. 𝐶𝑙𝑢𝑠𝑡𝑒𝑟𝑠add (𝐶𝑙𝑢𝑠𝑡𝑒𝑟𝑠; Data; zeros; 0); 𝑓𝑖𝑛𝑖𝑠ℎ𝑒𝑑0;
2. while finished==0 do 𝐶𝑙𝑠𝑆𝑖𝑧𝑒 𝐶𝑙𝑢𝑠𝑡𝑒𝑟𝑠:getSize/* The product of clusters */
3. 𝑃𝑎𝑟𝑡𝑖𝑡𝑖𝑜𝑛𝑖𝑛𝑔(Clusters; α); 𝑀𝑒𝑟𝑔𝑖𝑛𝑔(Clusters; α);
4. if |𝐿𝑎𝑟𝑔𝑒𝑠𝑡𝐶𝑙𝑢𝑠𝑡𝑒𝑟(𝐶𝑙𝑢𝑠𝑡𝑒𝑟𝑠)|<=α or 𝐶𝑙𝑢𝑠𝑡𝑒𝑟𝑠.getSize==𝐶𝑙𝑠𝑆𝑖𝑧𝑒 then 𝑓𝑖𝑛𝑖𝑠ℎ𝑒𝑑 _1
5. return [Clusters];
6. Procedure 𝑃𝑎𝑟𝑡𝑖𝑡𝑖𝑜𝑛𝑖𝑛𝑔 (𝐶𝑙𝑢𝑠𝑡𝑒𝑟𝑠; α)
7. U 𝐿𝑎𝑟𝑔𝑒𝑠𝑡𝐶𝑙𝑢𝑠𝑡𝑒𝑟(𝐶𝑙𝑢𝑠𝑡𝑒𝑟𝑠); while |U: objects|>α do 𝑤  𝑢𝑠𝑖𝑛𝑔 𝑒𝑞. (1);
8. 𝑖𝑓 (𝑤 == 0) 𝑡ℎ𝑒𝑛 U.𝑛𝑜𝑛𝑃𝑎𝑟𝑡𝑖𝑡𝑖𝑜𝑛𝑒𝑑(1); upgrade (𝐶𝑙𝑢𝑠𝑡𝑒𝑟𝑠;U);
9. resume; < 𝑡𝑚𝑝𝐶𝑙𝑢𝑠𝑡𝑒𝑟𝑠>Algorithm 1(𝑈, 𝑤);
10. if 𝐶𝑙𝑢𝑠𝑡𝑒𝑟𝑁𝑢𝑚 (𝑡𝑚𝑝𝐶𝑙𝑢𝑠𝑡𝑒𝑟𝑠)>1 then remove (Clusters; U);
11. attach (Clusters; 𝑡𝑚𝑝𝐶𝑙𝑢𝑠𝑡𝑒𝑟𝑠); U 𝐿𝑎𝑟𝑔𝑒𝑠𝑡𝐶𝑙𝑢𝑠𝑡𝑒𝑟(Clusters);
12. else 𝑤  𝑤 − (𝑤 ∗ 0.1); pass to step 21
13. 𝑃𝑟𝑜𝑐𝑒𝑑𝑢𝑟𝑒 Merging (Clusters; α)
14. 𝑀𝑒𝑟𝑔𝑖𝑛𝑔𝐿𝑖𝑠𝑡/* list of 𝑡𝑢𝑝𝑙𝑒𝑠<*//* 𝑐ℎ𝑖𝑙𝑑𝐶𝑙𝑢𝑠𝑡𝑒𝑟𝐼𝐷, 𝑝𝑎𝑟𝑒𝑛𝑡𝐶𝑙𝑢𝑠𝑡𝑒𝑟𝐼𝐷>*/
15. for each U in Clusters do j using eq. (2) and eq. (3); /* ID of cluster contained U
*/
16. if j not equal 0 then put<U:getID; j>in 𝑀𝑒𝑟𝑔𝑖𝑛𝑔𝐿𝑖𝑠𝑡;
17. while 𝑀𝑒𝑟𝑔𝑖𝑛𝑔𝐿𝑖𝑠𝑡 not similar to f produce
18. for each 𝑡𝑢𝑝𝑙𝑒 in 𝑀𝑒𝑟𝑔𝑖𝑛𝑔𝐿𝑖𝑠𝑡 do
19. < 𝑖; 𝑗>𝑡𝑢𝑝𝑙𝑒; if !𝑖𝑠𝑃𝑎𝑟𝑒𝑛𝑡(𝑀𝑒𝑟𝑔𝑖𝑛𝑔𝐿𝑖𝑠𝑡; i) then 𝑀𝑒𝑟𝑔𝑒𝐶𝑙𝑢𝑠(Clusters; 𝑖; 𝑗);
20. delete 𝑡𝑢𝑝𝑙𝑒 from 𝑀𝑒𝑟𝑔𝑖𝑛𝑔𝐿𝑖𝑠𝑡;
21. for i=1 to N do 𝑆𝑊𝑖′=𝑛𝑜𝑟𝑚(𝑆𝑊𝑖); 𝑆𝑠𝑖𝑚=∅;
22. for j=1 to M do 𝑆𝑖, 𝑤𝑠= 𝑤𝑠𝑆𝑖𝑚𝑊𝑆𝑖,, ;
23. if 𝑆𝑖, 𝑤𝑠>𝜀𝑎𝑛𝑑𝑷𝑢, ≠∅ then add US𝑢, into 𝑆𝑠𝑖𝑚;
24. end if
25. end for
26. if<𝑁𝑢𝑚 then//𝑁𝑢𝑚 is a threshold number Find the top-10 similar users 𝑈𝑠𝑖𝑚that have
used US𝑐,;
27. 𝑃𝑢,𝑖=𝜔
∑ 𝑈𝑆𝑢,𝑗∈𝑆𝑠𝑖𝑚
𝑆𝑖,𝑗
𝑤𝑠
𝑋 𝑃𝑢𝑗
𝑆𝑖,𝑗
𝑤𝑠 + 𝜔
∑ 𝑢𝑘 𝜖 𝑈𝑠𝑖𝑚
𝑆𝑢,𝑢𝑘
𝑢𝑠𝑒𝑟
𝑋 𝑃𝑢,𝑘𝑖
∑ 𝑢𝑘𝜖 𝑈 𝑆𝑖𝑚𝑆𝑢,𝑢𝑘
𝑢𝑠𝑒𝑟
28. else
29. 𝑃𝑢,𝑖=𝜔
𝑆𝑖,𝑗
𝑤𝑠
𝑋 𝑃𝑢𝑗
𝑆𝑖,𝑗
𝑤𝑠

1575
30. end if 𝑈𝑢,=SW𝑖 ′ × 𝑷𝑢,;
31. end for
32. return 𝑈𝑢, 1, 𝑈𝑢, 2, ⋯, 𝑈𝑢, 𝑁;
Algorithm 2: Web service graph construction
Input: 𝑆1 𝐻,𝑆2 𝐻,⋯,𝑆𝑁𝐻; 𝑆1 𝑃,𝑆2 𝑃,⋯,𝑆𝑁𝑃; 𝑈𝑢,1,𝑈𝑢,2,⋯,𝑈𝑢,𝑁; 𝜃𝐻, 𝜃𝑃, 𝛼, 𝛽, 𝛾; Web service
Graph 𝐺=(𝑉,), parameter 𝜆, adjacency matrix M
Output: Web service Graph 𝐺=(𝑉,); M set a of b ranked Web services
1. 𝑉=∅, 𝐸=∅;
2. for i=1 to N do
3. if 𝑆𝑖𝐻≥𝜃𝐻 or 𝑆𝑖𝑃≥𝜃𝑃 then
4. add to 𝑉;
5. end if
6. end for
7. or each node in 𝑉 do
8. S𝑐𝑜𝑟𝑒𝑢,=𝛼𝑆𝑖𝐻 +𝛽𝑆𝑖𝑃 +𝛾𝑈𝑢,;
9. end for
10. for each pair of nodes 𝑣𝑖 and 𝑣 in 𝑉 do
11. if (US𝑖,US𝑗)≥𝜏 then
12. add edge (𝑣𝑖,) to 𝐸;
13. end if
14. end for
15. return 𝐺=(𝑉,);
16. a=∅;
17. while|a|≤b do
18. find 𝑣𝑚𝑎𝑥=𝑎𝑟𝑔 max 𝑣∈(𝑉−A)(1−𝜆)𝑆𝑐𝑜𝑟𝑒𝑣+𝜆b|𝑁𝑣−𝑁(A)|;
19. a=a∪{𝑣𝑚𝑎𝑥 };
20. end while
21. return A;
An internet service diagram, 𝐺=𝑉, is an undirected weighted chart comprising of a lot of nodes 𝑉 and
a lot of edges 𝐸, wherein a hub indicates an Internet service competitor, i.e., 𝑣𝑖=US𝑖, and an edge signifies that
the associated nodes are comparative. 𝑉=𝐾 is the number of nodes (i.e., Web services) that appear in the
diagram. Be that as it may, here not all the internet services in the Internet service pool are utilized for
developing the Internet service diagram. Just the Internet services with a specific pertinence to client interests
are utilized.
2.4. Clustering quality
Clustering algorithms can be difficult to determine. False-positive and false-negative decisions are
penalized by external evaluation measures such as the rand index. A high number of output classes and output
clusters will compromise the quality of the output. Particularly concerning is the lack of a predetermined K-
value in the tested graph-based clustering algorithms. Pure measures penalize when the number of output
clusters exceeds the number of class labels, so it is preferable over other evaluation measures for reducing
quality trade-offs. The purity was calculated for each of the tested corpora and algorithms by
𝑝𝑢𝑟𝑖𝑡𝑦(𝐶, 𝑀) =
1
𝑁
∑ max
𝑗
𝑘
|𝑐𝑘 ∩ 𝑚𝑗|
(2)
with 𝑁 as the total number of documents, the set of clusters 𝐶 and 𝑀 as the set of classes.
3. RESULTS AND DISCUSSION
To perform reliable examinations, it is required to utilize huge scale true web services. To overcome
the tedious job of gathering and getting ready such data, Lausen and Haselwanter [11] share an enormous scale
genuine web services dataset gathered throughout their Distributed Reliability Assessment Mechanism for
Web services (WS-DREAM) test. WS-DREAM exists in a web creeping motor that shakes the openly
accessible Web Services Description Language (WSDL) document path of the web. It furthermore assumes
non-functional traits (e.g., QoS) of these web services, considered by 640 appropriated PCs situated in 25
unique nations, from Planet-Lab4. As a result, we get the top k-web hosting list which is recommended by the
system [21]-[25]. The following figures show exceptionally practical results for the proposed work clustering-
based web service recommendation (CWSR). Figure 2 shows a comparison of precision of Web service
Discovery (WSD) and CWSR approach and Figure 3 shows an F-score comparison between the existing
systems and the proposed one.

 ISSN: 2088-8708
1576
In this test, we executed 500 experiments to evaluate the pick time of our technique. The plans change
long; the range in our examination begins with plans comprising of 40 occupations to plans to contain 400
employments. Figure 4 demonstrates the determination time for the analysis of the clustering-based
methodology. The time to choose web hosting for each activity inside an arrangement is somewhere in the
range of 0.5 and 1.3 seconds.
Figure 2. Measurement of precision of WSD and our
method
Figure 3. Comparison of F-Score values for the
proposed system
Figure 4. Selection time in clustering-based approach
4. CONCLUSION
In the paper, we present an effective cluster-based novel technical solution to improve web services,
by building up the exhibition of serving suggestions. We have identified the use of clustering to investigate a
set of data in the real world. Data clustering can be used in the real world to investigate a set of data. Comparing
novel graph-based clustering algorithms with well-known vector-based algorithms is presented here. Our study
looked at how well each algorithm clustered and how it performed in general. In contrast to the graph-based
clustering algorithm, the vector-based algorithms performed more efficiently. The classical approach of using
the k-means algorithm, however, requires the user's intervention a priori, which eliminates use cases where the
user can investigate the input data before that is applied to the clustering. However, the graph-based clustering
algorithm demonstrates that a good categorization can be achieved even without requiring the pre-requisite k-
value. The distance between search points is reduced by our method. Furthermore, our suggested strategy offers
greater accuracy than existing systems, as shown by the results of the tests. A small number of clusters (k=6)
results in a large memory cost and a long run time for the k-means algorithm. The rough clustering algorithm

1577
improves all three criteria concerning rough clustering. As a result of the need to estimate upper and lower
approximations during the process of updating a new focus, there is the creation of a high average error cost.
ACKNOWLEDGMENT
The authors wish to express their appreciation to the Ministry of Education and Training for
supporting this research project as part of the Ministerial Program of Science and Technology CTB.2021.DNA.
“Research on applying deep learning model to recognize ripe pineapple period in Quang Nam-Da Nang”.
REFERENCES
[1] H. Wang, J. Z. Huang, Y. Qu, and J. Xie, “Web services: Problems and future directions,” Journal of Web Semantics, vol. 1, no. 3,
pp. 309–320, Apr. 2004, doi: 10.1016/j.websem.2004.02.001.
[2] M. J. Mokarrama and M. S. Arefin, “RSF: A recommendation system for farmers,” in 2017 IEEE Region 10 Humanitarian
Technology Conference (R10-HTC), Dec. 2017, pp. 843–850, doi: 10.1109/R10-HTC.2017.8289086.
[3] M. J. Pazzani and D. Billsus, “Content-based recommendation systems,” in The Adaptive Web, Berlin, Heidelberg: Springer Berlin
Heidelberg, pp. 325–341.
[4] A. Kumar, S. Sarkar, and C. Pradhan, “Recommendation system for crop identification and pest control technique in agriculture,”
in 2019 International Conference on Communication and Signal Processing (ICCSP), Apr. 2019, pp. 0185–0189, doi:
10.1109/ICCSP.2019.8698099.
[5] C. Atkinson, P. Bostan, O. Hummel, and D. Stoll, “A practical approach to web service discovery and retrieval,” in IEEE
International Conference on Web Services (ICWS 2007), Jul. 2007, pp. 241–248, doi: 10.1109/ICWS.2007.12.
[6] S. K. S. Raja, R. Rishi, E. Sundaresan, and V. Srijit, “Demand based crop recommender system for farmers,” in 2017 IEEE
Technological Innovations in ICT for Agriculture and Rural Development (TIAR), Apr. 2017, pp. 194–199, doi:
10.1109/TIAR.2017.8273714.
[7] C. Platzer, F. Rosenberg, and S. Dustdar, “Web service clustering using multidimensional angles as proximity measures,” ACM
Transactions on Internet Technology, vol. 9, no. 3, pp. 1–26, Jul. 2009, doi: 10.1145/1552291.1552294.
[8] N. Kokash, “A comparison of web service interface similarity measures,” Frontiers in Artificial Intelligence and Applications,
vol. 142, 2006.
[9] K. Elgazzar, A. E. Hassan, and P. Martin, “Clustering WSDL documents to bootstrap the discovery of web services,” in 2010 IEEE
International Conference on Web Services, Jul. 2010, pp. 147–154, doi: 10.1109/ICWS.2010.31.
[10] Y. Liu, A. H. Ngu, and L. Z. Zeng, “QoS computation and policing in dynamic web service selection,” in Proceedings of the 13th
international World Wide Web conference on Alternate track papers and posters-WWW Alt. ’04, 2004, Art. no. 66, doi:
10.1145/1013367.1013379.
[11] H. Lausen and T. Haselwanter, “Finding web services,” The 1st European Semantic Technology Conference, 2007.
[12] H. Ma, I. King, and M. R. Lyu, “Effective missing data prediction for collaborative filtering,” in Proceedings of the 30th annual
international ACM SIGIR conference on Research and development in information retrieval-SIGIR ’07, 2007, Art. no. 39, doi:
10.1145/1277741.1277751.
[13] Y. Wang and J. Vassileva, “Towards trust and reputation based web service selection,” In Multi- Agent and Grid Systems (MAGS)
Journal, 2007.
[14] B. T. G. S. Kumara, I. Paik, W. Chen, and K. H. Ryu, “Web service clustering using a hybrid term-similarity measure with ontology
learning,” International Journal of Web Services Research, vol. 11, no. 2, pp. 24–45, Apr. 2014, doi: 10.4018/ijwsr.2014040102.
[15] E. M. Maximilien and M. P. Singh, “Multi agent system for dynamic web services selection,” in Proceedings of 1st Workshop on
Service-Oriented Computing and Agent-Based Engineering (SOCABE at AAMAS), 2005, pp. 25–29.
[16] E. M. Maximilien and M. P. Singh, “Reputation and endorsement for web services,” ACM SIGecom Exchanges, vol. 3, no. 1,
pp. 24–31, Dec. 2001, doi: 10.1145/844331.844335.
[17] H. Sofian, J. T. Chia Ming, S. Muhammad, and N. Mohd Noor, “Calcification detection using convolutional neural network
architectures in Intravascular ultrasound images,” Indonesian Journal of Electrical Engineering and Computer Science, vol. 17,
no. 3, pp. 1313–1321, Mar. 2020, doi: 10.11591/ijeecs.v17.i3.pp1313-1321.
[18] O. Feddaoui, R. Toufouti, L. Jamel, and S. Meziane, “Fuzzy logic control of hybrid systems including renewable energy in
microgrids,” International Journal of Electrical and Computer Engineering (IJECE), vol. 10, no. 6, pp. 5559–5569, Dec. 2020, doi:
10.11591/ijece.v10i6.pp5559-5569.
[19] C. S. Baidal, N. X. Arreaga, and V. S. Padilla, “Design and testing of a dynamic reactive signage network towards fire emergency
evacuations,” International Journal of Electrical and Computer Engineering (IJECE), vol. 10, no. 6, pp. 5853–5860, Dec. 2020,
doi: 10.11591/ijece.v10i6.pp5853-5860.
[20] B. E. Elnaghi, M. E. Dessouki, M. N. Abd-Alwahab, and E. E. Elkholy, “Development and implementation of two-stage boost
converter for single-phase inverter without transformer for PV systems,” International Journal of Electrical and Computer
Engineering (IJECE), vol. 10, no. 1, pp. 660–669, Feb. 2020, doi: 10.11591/ijece.v10i1.pp660-669.
[21] R. Burke, “Hybrid web recommender systems,” in The Adaptive Web, Berlin, Heidelberg: Springer Berlin Heidelberg,
pp. 377–408.
[22] H. H. C. Nguyen, A. T. Luong, T. H. Trinh, P. H. Ho, P. Meesad, and T. T. Nguyen, “Intelligent fruit recognition system using deep
learning,” in Lecture Notes in Networks and Systems, Springer International Publishing, pp. 13–22, 2021.
[23] H. H. C. Nguyen, D. H. Nguyen, V. L. Nguyen, and T. T. Nguyen, “Smart solution to detect images in limited visibility conditions
based convolutional neural networks,” in Advances in Computational Collective Intelligence, Springer International Publishing,
pp. 641–650, 2020.
[24] Papaioannou, I., Tsesmetzis, D., Roussaki, I., andAnagnostou, M. (2006). “A QoS Ontology Language forWeb-Services”
In Proc. of 20th Intl. conf. on Advanced Information Networking and Applications (AINA), vol. 1, Vienna, Austria.
[25] Y. Lai and J. Zeng, “A cross‐language personalized recommendation model in digital libraries,” The Electronic Library, vol. 31,
no. 3, pp. 264–277, May 2013, doi: 10.1108/EL-08-2011-0126.

 ISSN: 2088-8708
1578
BIOGRAPHIES OF AUTHORS
Ha Huy Cuong Nguyen obtained his doctorate in Computer Science/Resource
Allocation Cloud Computing in 2017 from the University of Danang. He has published over 50
research papers. His main research interests include the resource allocation, detection,
prevention, and avoidance of cloud computing and distributed systems. He serves as a technical
committee program member, track chair, session chair and reviewer of many international
conferences and journals. He is a guest editor of “International Journal of Information
Technology Project Management (IJITPM)” with Special Issue On: Recent Works on
Management and Technological Advancement. Currently, he is working at Software
Development Centre, The University of Danang. He can be contacted at
email:nhhcuong@vku.udn.vn.
Bui Thanh Khiet acquired his Master's degree from Posts and Telecommunications
Institute of Technology in Ho Chi Minh in 2012. He is working at Institute of Engineering
Technology, Thu Dau Mot University as a lecture. At present, he is a Ph.D. student at Computer
Science, Faculty of Computer Science and Engineering, Ho Chi Minh City University of
Technology (HCMUT), VNUHCM. His research focuses on Cloud computing. He can be
contacted at email: khietbt@tdmu.edu.vn.
Van Loi Nguyen received his Master of Engineering in Computer Science from
the University of Danang, Vietnam in 2010 and Ph.D. degree from Soongsil University, Korea
in 2017. He is currently a lecturer at Vietnam-Korea University of Information and
Communication Technology, the University of Danang. His research interests include
multimedia, information retrieval, artificial intelligence, database, and IoT. He can be
contacted at email: nvloi@vku.udn.vn.
Thanh Thuy Nguyen received the Engineer degree of the computing in 1982 and
the Ph.D. degree of the computer science in 1987 from the Hanoi University of Science and
Technology, Hanoi, Vietnam. He had been a Professor of computer science with the Hanoi
University of Science and Technology until 2011 and since then has been with the VNU
University of Engineering and Technology. Prof. Nguyen Thanh Thuy is now Head of the Key
Laboratory for Artificial Intelligence (AI) at VNU UET. He is also Deputy Director of the
National R&D Program KC4.0/2019-2025 (Ministry of Science and Technology).
His AI research interests are mainly in Knowledge systems, Soft-computing, Data mining,
Machine learning, and Hybrid Intelligent systems. He can be contacted at email:
nguyenthanhthuy@vnu.edu.vn.

An effective method for clustering-based web service recommendation

More Related Content

Similar to An effective method for clustering-based web service recommendation (20)

More from IJECEIAES (20)

Recently uploaded (20)

An effective method for clustering-based web service recommendation