HeLi1
-
(School of Finance and Trade Management, Chengdu Industry and Trade College, Chengdu
611731, China)
Copyright © The Institute of Electronics and Information Engineers(IEIE)
Keywords
Infant and toddler care, Intelligent management system, Multi neighborhood perception, , Service quality perception, Time perception, Heterogeneous environment
1. Introduction
Information management has become more and more popular in daily life and work. In
order to meet the growing needs of people, various cloud server platforms provide
users with a large number of Web services with different functions [1,2]. Web services are a concept and technology that turns the Internet into a virtual
computing environment, in which users can use any type of client software to call
web services. Faced with a massive amount of web service information, how users can
quickly and accurately find services that meet their needs to be solved [3]. To address the above issues, Web service recommendation algorithms have emerged,
mainly by analyzing the historical information of users calling services to recommend
services that meet their needs [4]. Therefore, the current recommendation field mainly relies on QoS for web service
recommendation. Users can quickly distinguish numerous services with similar functions
and accurately call services that meet their needs [5,6].
However, there are still some important problems to be solved in the development of
Web service recommendation. For example, the application of edge computing, users
may call services on different terminals, different edge servers, and different service
platforms contain different information of users, resulting in user data distributed
in heterogeneous environments [7,8]. Therefore, how to perform service recommendation in heterogeneous environments has
become a hot research topic for scholars. Service recommendation in heterogeneous
environments is mainly divided into scenarios with smaller single platform data scale
and scenarios with larger cross platform data scale [9]. For application scenarios with small data scales on a single platform, most recommendation
algorithms currently analyze user behavior characteristics and only use user service
interaction information as the algorithm dataset for web service recommendation, without
considering the impact of external factors such as location and time on service recommendation.
These algorithms cannot solve the problem of dynamic QoS values generated by user
service calls. For application scenarios with large cross platform data scale, since
the QoS data of users is relatively scattered, edge computing needs to be applied
to gradually increase the data of each platform to relieve the pressure of data storage
[10,11]. However, although the existing algorithms introduce the concept of edge computing,
they are not really applied in the process of service prediction. At the same time,
when dealing with QoS data from users across platforms, there is a risk of user privacy
leakage [12,13]. Therefore, service recommendations in cross platform scenarios and privacy protection
of user data are also the focus of research work. In summary, web service recommendations
in the era of big data and intelligent information are suitable for heterogeneous
environments and are based on QoS for web service recommendations [14].
Therefore, how to recommend web services based on QoS in heterogeneous environments
is a major challenge in current research work. This article proposes a time aware
recommendation algorithm for recommendation scenarios with small data scales on a
single platform in heterogeneous environments. This algorithm introduces a time dimension
into the dataset to add time aware factors, making user interaction data dynamic and
no longer limited to the dimensions of users and services [15,16]. Not only does it improve the accuracy of service recommendations, but it also meets
the dynamic and real-time requirements of service invocation. A privacy preserving
cross platform data feature aware service recommendation algorithm is proposed for
recommendation scenarios with large data scales on multiple platforms in heterogeneous
environments [17]. This algorithm integrates historical QoS data of users from multiple platforms for
comprehensive analysis, while protecting their personal privacy data. It breaks through
the limitations of single platform data in a single edge server and grasps more connections
between users and services. Not only does it implement web service recommendation
in a cross-platform environment, but it also alleviates the cold start problem by
adopting model-based CF, further improving the accuracy and security of service recommendation
[18]. This paper designs and implements a road prediction recommendation system in smart
transportation, and promotes and applies the proposed web service recommendation algorithm
in heterogeneous environments to smart transportation scenarios. Users can real-time
grasp the traffic situation of the road in this system, and the system plans and recommends
reasonable paths for users to solve the problem of traffic congestion in daily life.
Collaborative filtering (CF) algorithm, as a traditional service recommendation algorithm,
mainly predicts missing QoS values Hybrid-based CF predicts missing QoS values by
weighting. When using a model-based collaborative filtering algorithm, the current
prediction model is trained by extracting a small portion of data from the real dataset
as the training set [19,20].
2. Relevant Theoretical Foundations
2.1. Collaborative Filtering Recommendation Algorithm
The user-based CF algorithm starts from the user's perspective, analyzes user historical
behavior data to discover user preferences, divides users based on user preferences,
finds similar users with similar preferences, as shown in Eqs. (1) and (2), and finally recommends services that the user has not called but has similar preferences
to the target user. The nearest neighbor search calculates the similarity between
users, The core of CF algorithm is similarity calculation.
Firstly, identify the user groups who have both called services $i$ and $j$, and then
calculate the similarity between users based on their behavior ratings. The similarity
calculation generally involves the following methods. As shown in Eqs. (3) and (4). It indicates that the cosine similarity is larger, which means that the similarity
between users is higher.
When calculating similarity between users, compared to cosine similarity, this method
adds the average score of users to adjust for each independent rating, as shown in
Eqs. (5) and (6), reducing the impact of individual user rating biases on the results.
If two users have not called the same service, that is, if the intersection of services
I is an empty set, it indicates that the similarity between these two users is also
zero, indicating that there is no similarity between the two. In the recommendation
process, similarity calculation mainly refers to a non-negative number, as shown in
Eqs. (7) and (8). Generally, similarity calculation sets the results between $[1$, $1]$ or $[0 $,
$1]$. According to the needs, the above methods are selected to complete the similarity
calculation between users. Then, the TopK algorithm is used to obtain the first $K$
similar users of the user as a set of similar users, and a weighted sum is used for
service prediction. The principle of the service-based CF algorithm is similar to
that of the user-based CF algorithm, but only the service itself is used when searching
for neighborhoods. Relationship refers to finding similar services based on historical
preferences, user has not yet called but may prefer, and obtain a descending list
of services as recommended services.
When calculating service neighborhoods in service-based CF, as shown in Eqs. (9) and (10), and then predict the services that the current user has not yet called but may prefer
based on their historical preferences. Use a weighted sum method to obtain the prediction
results, predict and complete the data matrix.
2.2. Potential Factor Model
User based CF and service-based CF differ in application scenarios. User based CF
recommends based on user similarity, taking into account the diversity of the system
and making it more social. It allows users to quickly find their ``friends'' and dynamically
update their ``self'' recommendation list based on the services called by ``friends'',
as shown in Eqs. (11) and (12). Therefore, the recommendation system can provide users with more diverse choices.
The service-based CF algorithm, on the other hand, looks for a similar set of services
that users have called from a single service perspective, making it more suitable
for applications with stable changes in interests. Users are more inclined to search
for a type of service within a certain period of time, therefore, the diversity of
service recommendations is not as good as user-based CF.
However, user-based CF always tends to recommend popular services, causing serious
``head problem'', while service-based CF has better novelty in recommendations and
is good at considering unpopular recommendations. Memory based CF, as shown in Eqs.
(13) and (14), essentially reads all user data into memory and performs corresponding calculations.
It is suitable for scenarios with small-scale datasets, but when the data size is
particularly large, this method will no longer be applicable.
In this situation, model-based CF is needed. It is based on the preferences of the
target user, relies on machine learning models, obtains a recommendation model in
advance through offline training, and then predicts and recommends based on real-time
user preferences. For example, in the data matrix of QoS values left by user service
calls, as shown in Eqs. (15) and (16), has the advantages of alleviating cold start, saving running space, fast training
and prediction speed, and avoiding overfitting to help compensate for the shortcomings
of memory-based CF.
Which has inherent rich correlations between data. The generated matrix is usually
approximated as a low rank matrix product with numerical values in both rows and columns.
The latent factor model is considered the latest technology in recommendation systems,
usually using dimensionality reduction methods to predict missing values in matrices.
The basic idea of dimensionality reduction method is equivalent to rotating the coordinate
axis, as shown in Eqs. (17) and (18), to eliminate pairwise correlations between dimensions and robustly predict dimensionality
reduction, rotation, and fully specified representations from incomplete data matrices.
Points that do not belong to the core point but are within the neighborhood of other
core points are called boundary points. As shown in Eqs. (19) and (20), point is called noise points. Select any object $p$ from the dataset, and if object
$p$ is not processed, if $q$ is not divided into other clusters, then the object $q$
is divided into cluster $C$.
3. Time Aware Service Recommendation Algorithm Based on Continuous Time Slices
3.1. Design of Time Aware Service Recommendation Algorithm Based on Continuous Time
Slices
Faced with it has become a challenge for users to quickly and accurately find the
required services. The traditional service recommendation algorithm uses neighborhood
of similar user sets for target users or to predict and complete missing by searching
for similar service sets for target services [21,22]. However, neighborhood-based CF algorithms often highlight their advantages in datasets
with relatively small data volumes. So, when using this method for service recommendation
in small data scenarios, problems such as cold start often arise [23,24]. Fig. 1 shows the overall flowchart of the multi neighborhood perception optimization algorithm.
To address this challenge, a latent factor model based on data feature perception
is proposed. The model first extracts the sparse dataset through the latent factor
$f$, then clusters and partitions the extracted LF matrix using each neighborhood
matrix. Finally, the original data matrix is integrated in the order of partitioning
[25,26].
Fig. 1. Overall flowchart of multi neighborhood aware optimization algorithm.
In addition, to address the challenge of cold start and assist users in executing
high-precision service recommendations, a flexible matrix factorization technique
is proposed for future time gaps [27,28]. With the popularization of the Internet services. In order to provide users with
optimal services in a distributed environment, a federated learning method based on
location interest point recommendation is proposed. By optimizing the interest points
proposed through matrix factorization, the optimal cloud services are recommended
to users [29,30]. Fig. 2 shows the flowchart of the resource allocation optimization algorithm. A high-precision
and high-performance recommendation algorithm is designed, and a rough multidimensional
matrix model based on collaborative filtering algorithm is proposed. Standards that
are helpful for service selection are designed based on user and service attributes.
Fig. 2. Resource allocation optimization algorithm flowchart.
Recommendation systems require collecting large amounts of historical QoS data, which
poses potential privacy risks. To address this, a similarity-based privacy protection
strategy was designed to protect user privacy while ensuring data availability. A
location-based low-rank matrix factorization algorithm was proposed, incorporating
latitude and longitude information for scalability and flexibility. However, this
model struggles to integrate contextual user and service features, limiting its performance
in large-scale applications. In response, deep learning is utilized to mine sparse
user data for more valuable insights. A personalized service recommendation algorithm
based on a long short-term neural network with location data is introduced, employing
a dual-layer attention mechanism to capture users' long- and short-term sequences.
To further enhance service quality and improve user interaction, a secure collaborative
filtering recommendation algorithm integrating similar content is proposed, utilizing
AI technology to optimize physical space and improve service recommendations.
Fig. 3. Risk prediction and management algorithm flowchart.
3.2. Implementation of TASQP Time Aware Service Recommendation Algorithm
By seamlessly integrating similarity information between extracted web services and
users into neural networks, fast and accurate predictions can be achieved. Apply Graph
Neural Network (GNN) to recommendation systems, while utilizing the interaction, to
integrate into the GNN for prediction. Although deep learning models have been applied
to data mining and interpretation of recommendation system neighborhoods, and their
practicality and accuracy have been proven, there are not many complete applications
of deep learning models in recommendation systems. Therefore, a deep neural network
integration method was proposed, which combines neighborhood based and neural network-based
models, obtains outputs from these two models, and then combines them with another
different neural network model. Fig. 4 shows the evaluation of the number of infants and young children. A new service combination
recommendation method is proposed, which mainly consists of two steps: the first step
is to establish a service component based on recursive neural networks, and use RNN
algorithm to recommend other matching services to users, forming a service combination.
The second step is based on Bayesian service combination clustering. In the process
of forming user combinations, considering the diversity of users, Bayesian clustering
is used to model their interests, and then recommend service combinations that meet
their needs to users.
Fig. 4. Assessment of infant and toddler population.
Overall, using deep learning models can integrate information across platforms and
improve prediction accuracy to a certain extent. However, model training is a very
time-consuming process and has poor interpretability, which does not meet the real-time
requirements of users for web services. Traditional service recommendation algorithms
only consider the behavior data of users calling services, while ignoring the impact
of user location and time factors on the accuracy of QoS prediction. Recommendation
systems often assume that user behavior is static and unique, without considering
the impact of dynamic data on QoS. With the deepening of research, some scholars have
proposed service recommendation algorithms based on time perception, and have verified
through experiments that this algorithm has greatly improved the accuracy of recommendation
results. Fig. 5 shows the assessment of childcare service needs. It is found that traditional FM
ignores the long-term dependency relationship between users and services that dynamically
changes over time. To solve this problem, they proposed a new model that combines
gated recurrent units with self-attention (SAGRU) and projection decomposition machine
(PFM). By assigning different weights to self-attention, the tensor form of user service
time is extracted, and a bottleneck structure is designed in the output to compress
encoding and filter noise. A two-stage service quality prediction method based on
time perception was proposed. The first stage is to predict the missing QoS value
based on historical time slices.
Fig. 5. Assessment of demand for childcare services.
Needs of edge services in different time periods. By weighting and synthesizing the
service features of different dimensions in different time periods, user location
with traditional collaborative filtering algorithms. In addition, in order to solve
the problem of users being able to quickly obtain high-quality cloud services in the
cloud environment, an improved multi-layer perceptron algorithm was studied and designed
to recommend the best and reliable cloud services to users. This not only considers
the impact of the user's location and time to the virtual machine on the efficiency
of selecting cloud services, but also takes into account the location of the virtual
machine and the efficiency of cloud service data transmission. In order to introduce
time factors into the dataset, Fig. 6 shows the evaluation chart of parental satisfaction survey. A novel algorithm was
designed to segment the dataset from three dimensions and predict missing QoS values.
Finally, the data from the three dimensions was weighted and integrated to obtain
the final prediction result. This algorithm not only considers time perception but
also greatly improves the accuracy of prediction. Overall, time aware web service
recommendation algorithms introduce a temporal dimension to make service prediction
dynamic, but there are certain limitations on data scale in application scenarios.
Fig. 6. Parent satisfaction survey evaluation chart.
4. Intelligent Management of Infant and Toddler Care Services Based on Multi Neighborhood
Perception Optimization Algorithm
Collaborative filtering is a classic algorithm that utilizes collective intelligence.
Different levels of service and record them, in order to help other users filter service
information in the later stage and achieve the purpose of prediction and recommendation.
Collaborative filtering, with its excellent filtering speed and robustness, the other
is model based collaborative filtering recommendation algorithms. In heterogeneous
service recommendation scenarios, users may call Web services in a single platform
environment or in a cross-platform environment. This chapter focuses on the single
platform service recommendation algorithm in heterogeneous environments. The model-based
collaborative filtering algorithm collects known QoS values left by users calling
services, and uses various models in deep learning or machine learning to train and
predict the services that users may call. Table 1 is experimental dataset description. In application scenarios with small data scales
on a single platform in heterogeneous environments, memory-based CF recommendation
algorithms are generally used. However, most memory-based CF algorithms currently
do not rely on other additional information and only rely on the interaction data
of users calling services, that is, only considering the two dimensions of users and
services. So only considering the user and service dimensions ignores the influence
of time on user personalization and service popularity. And currently, the existing
time-based service recommendation algorithms only introduce the time dimension into
the dataset, without considering the interaction data connections between time slices.
Table 1. Experimental dataset description.
|
Number of users
|
339
|
|
Number of services
|
5825
|
|
Number of service calls
|
1974675
|
|
Number of countries where services are located
|
73
|
|
The number of countries that the user belongs to
|
30
|
|
The average response time
|
0.81s
|
|
Average throughput
|
44.03s
|
Therefore, in order to consider the impact of time factors on service prediction in
heterogeneous single platform service scenarios, this chapter proposes a time aware
web service recommendation algorithm based on continuous time slices. This algorithm
introduces time elements into the traditional user and service dimensions, and predicts
the user's call to the service on the current time slice based on the QoS value left
by the users. Not only considering the QoS value of the user on the current time slice,
but also referring to the QoS values on the first T time by assigning different weights
to the first T time slices. Not only does it meet the dynamism of user service invocation,
but it also improves the accuracy of service recommendations. The specific work includes
adding a time dimension to the dataset to introduce time perception, researching web
and designing a neighborhood-based CF service recommendation framework to make web
service recommendations dynamic. Fig. 7 shows the service response time evaluation graph, and then predicts the missing QoS
values. When users access cloud services, the concept of time perception is introduced,
and a time dimension is added to the user service dimension of the original two-dimensional
dataset to consider the situation of calling services at different time points. And
in real life, when recommending services to users, not only accuracy issues need to
be considered, but real-time performance is also a major challenge. This chapter proposes
a time aware service recommendation algorithm based on continuous time slices. This
method mainly consists of four steps. Firstly, it needs to preprocess the publicly
available dataset on WS-Deram, which introduces a time dimension and describes the
response time.
Fig. 7. Service response time evaluation chart.
This algorithm divides the data according to the time dimension and obtains QoS records
of user service calls on 64 time slices; Achieve web service recommendation for the
target user. In real service recommendation scenarios, the same service at different
points in time, and over time and changes in the network environment, it can lead
to changes in the QoS value left by users accessing the service. The TopK algorithm
is a method that can avoid sorting all data and only require partial sorting, meaning
that once the desired sorted data is obtained, sorting can be stopped. Due to the
simplicity and efficiency of the algorithm, this section also needs to rely on the
TopK algorithm to find the corresponding optimal similarity. The commonly used TopK
sorting algorithm in experiments is heap sorting, where the large root heap maintains
an ordered array of size $K$. Table 2 shows the comparison of MAE experimental data, where for the elements in the large
root pile, the root is the largest number. Afterwards, each time an element is taken
from the original array and compared with the elements in the large root heap are
still the top $K$ ranked number and the root element is still the largest number.
Repeat the above steps in this way until an array of length $K$ is generated.
Table 2. Comparison of MAE experimental data.
|
Method
|
Sparsity of data matrix
|
|
20%
|
25%
|
30%
|
35%
|
40%
|
Average value
|
|
Mae
|
|
Upcc
|
0.926
|
0.895
|
0.849
|
0.781
|
0.713
|
0.832
|
|
Ipcc
|
0.954
|
0.926
|
0.874
|
0.825
|
0.768
|
0.869
|
|
Uipcc
|
0.899
|
0.862
|
0.815
|
0.738
|
0.681
|
0.799
|
|
Tosem
|
0.758
|
0.735
|
0.714
|
0.689
|
0.623
|
0.703
|
|
Tasqp
|
0.722
|
0.702
|
0.687
|
0.653
|
0.589
|
0.671
|
According to the previous step, after completing the similarity calculation of all
users and services, the TopK algorithm is used to obtain the similarity matrix of
the obtained users and services, respectively, Utarget and the set of similar services
for the target service Starget. The similarity matrix is obtained by calculating the
Pearson correlation coefficient, and then the TopK algorithm is used to select the
top $K$ users as the set of similar users for the target users. These users and the
target users call a large number of the same services user may call. For the target
service, the algorithm TASQP also uses the same method to obtain the first $K$ similar
services. Note that $K$ is a variable and has different effects on different datasets.
In this section, parameters need to be adjusted through multiple experiments based
on the dataset used in the algorithm to obtain the optimal $K$ value. However, in
the dataset of real web service invocation scenarios, the user service data on each
time slot is very sparse, with a large number of missing values, and they may only
have a few similar users and services. Fig. 8 shows the health monitoring data graph, where the variable is a weight value. By
changing the value of the variable, On the overall prediction results is controlled.
That is to say, in real application scenarios, the QoS data on each time slot is constantly
changing. This section needs to measure the most suitable value through multiple experimental
data results to achieve the accuracy of the final prediction result.
Fig. 8. Health monitoring data graph.
5. Experimental Analysis
In real scenarios, it is unreasonable to rely solely on data from the current time
slot to make predictions when recommending services that meet the needs of the target
user at a certain point in time. In the face of massive services in cloud servers,
Fig. 9 shows the activity participation evaluation graph. Users may only call a small portion
of the services, and this data may also contain some untrusted data. It is inevitable
that there will be some abnormal users in the recommendation system who may intentionally
or unintentionally mislead the called services. If the interference of this untrusted
user information is not ruled out, no matter how the algorithm model is optimized,
it will have an impact on the results of service recommendation. So, the data that
can truly be used to recommend services to target users is very sparse.
Fig. 9. Activity participation evaluation chart.
In order to obtain more interactive data, this section found through careful observation
and research that the services called by users during a certain time period have similarities,
that is, users have certain regularity and correlation when calling services during
the same time period. Therefore, in order to obtain more valuable data, Fig. 10 shows the security event occurrence rate graph. When calling web services in real
scenarios, users often have noisy data in their QoS values, such as abnormal data
caused by malicious brushing or server failures.
Fig. 10. Security incident occurrence rate chart.
Users and services and filter out noisy data in the dataset, the accuracy of recommendations
is improved. In order to reduce errors caused by certain random factors, perform multiple
experiments on the DBSLF model and take the average as the final result. Fig. 11 shows the comparison between expense and revenue. The DBSLF model proposed in this
chapter adopts the SPP strategy to not only achieve privacy protection but also ensure
high data availability. Then, LF and DBSCAN models are used to predict missing values,
in addition, this section will compare the proposed DBSLF algorithm model with other
classic algorithms.
Fig. 11. Comparison of expense expenditure and income.
This algorithm improves the local sensitive hashing algorithm and uses PCC to establish
a hash user table. Then, based on the hash user table, it replaces the columns in
the data matrix to select user neighborhoods and perform QoS prediction. This algorithm
learns features from the user dimension and service dimension through the LFM model,
and predicts missing values based on the learned features. Fig. 12 shows the system performance indicator chart. The algorithm introduces a new method
of selecting service neighborhoods and incorporates the potential factor features
of services into matrix decomposition to predict missing values. This algorithm combines
geographic information such as longitude and latitude with trust mechanisms, and integrates
it into matrix decomposition models to predict missing values.
Fig. 12. System performance indicators chart.
6. Conclusion
This article explores the research of web service recommendation algorithms in heterogeneous
environments. This algorithm introduces a time dimension into the dataset, making
the originally static data dynamic. It not only improves the accuracy of service quality
prediction but also ensures the timeliness of recommended services to users, improving
the efficiency of recommendations. The second part proposes a cross-platform data
feature aware service recommendation algorithm DBSLF based on privacy protection.
This algorithm integrates the historical data of users calling services on different
edge server platforms while protecting user privacy, and pursues more comprehensive
and accurate recommendation results by integrating multi-platform data. Set the data
density of the data matrix to 20%, 25%, 30%, 35%, and 40%. For example, setting the
data density to 20% indicates that 20% of the data in the original time slice matrix
dataset is the test set to compare the predicted missing QoS values. Taking response
time as an example, the average MAE of this algorithm has decreased by 24.1%, 22.8%,
16.1%, and 4.6%, respectively.
In order to effectively solve the service recommendation problem in single platform
scenarios in heterogeneous environments, continuous time slices are proposed. Considering
the time factor, the process of service recommendation is dynamic. By referring to
historical time slices to predict missing service quality values in the current time
slice, in cross platform scenarios in heterogeneous environments, a privacy preserving
cross platform data feature aware service recommendation algorithm is proposed. Sharing
and integrating data from various platforms while protecting user privacy, using latent
factor models to solve cold start problems, and using density-based clustering and
matrix decomposition techniques to improve prediction accuracy. Finally, integrating
historical data of cross platform users to comprehensively utilize data feature information
and efficiency of service recommendations. Based on privacy protection proposed in
the second part, a road prediction recommendation system in smart transportation has
been comprehensively designed and implemented to alleviate the problem of road congestion
in daily travel.
REFERENCES
J. Peake, M. Amos, N. Costen, G. Masala, and H. Lloyd, ``PACO-VMP: Parallel Ant Colony
Optimization for Virtual Machine Placement,'' Future Generation Computer Systems,
prepublish, pp. 174--186, 2022.
O.~D. Montoya, L.~F. Grisales Noreña, L.~Barrios Alvarado, A.~Londoño Arias, and C.~Álvarez
Arroyo, ``Efficient Reduction in the Annual Investment Costs in AC Distribution Networks
via Optimal Integration of Solar PV Sources Using the Newton Metaheuristic Algorithm,''
Applied Sciences, vol.~11, no.~23, Article no.~11525, December 2021.
J.~O. Agushaka and A.~E. Ezugwu, ``Evaluation of several initialization methods on
arithmetic optimization algorithm performance,'' Journal of Intelligent Systems, vol.~31,
no.~1, pp.~70--94, 2021.
P.~Li, B.~Qiu, and J.~Chen, ``A method of reverse seeking material parameters by using
finite element optimization algorithm,'' Vibroengineering Procedia, Volume 39, pp.
58--62, December 2021.
J.~Wang, P.~Wang, B.~Li, R.~Fu, S.~Zhao, and H.~Zhang, ``Discriminative optimization
algorithm with global–local feature for LIDAR point cloud registration,'' International
Journal of Remote Sensing, vol.~42, no.~23, pp.~9003--9023, 2021.
M.~Irfan, S.-R. Oh, and S.-B. Rhee, ``An effective coordination setting for directional
overcurrent relays using modified harris hawk optimization,'' Electronics, vol.~10,
no.~23, Article no.~3007, December 2021.
Q.~Liu, Y.~Liu, C.~Zhang, Z.~Ruan, W.~Meng, Y.~Cai, and Q.~Ai, ``SEMG-based dynamic
muscle fatigue classification using SVM with improved whale optimization algorithm,''
IEEE Internet of Things Journal, vol.~8, no.~23, pp.~16835--16844, 2021.
A.~Sayed, M.~Ebeed, Z.~M.~Ali, A.~B.~Abdel-Rahman, M.~Ahmed, S.~H.~E.~Abdel Aleem,
A.~El-Shahat, and M.~Rihan, ``A hybrid optimization algorithm for solving of the unit
commitment problem considering uncertainty of the load demand,'' Energies, vol.~14,
no.~23, Article no.~8014, November 2021.
M.~H.~Nadimi-Shahraki, S.~Taghian, S.~Mirjalili, L.~Abualigah, M.~Abd Elaziz, and
D.~Oliva, ``EWOA-OPF: Effective whale optimization algorithm to solve optimal power
flow problem,'' Electronics, vol.~10, no.~23, Article no.~2975, December 2021.
H. S. Kaur, P. Rachana, R. P. de Prado, M. Wozniak, and P. D. Bidare, "Long short-term
memory network-based metaheuristic for effective electric energy consumption prediction,''
Applied Sciences, vol. 11, no. 23, 11052, 2021.
J. W. Yoon, S. Lim, and S. K. Park, ``Combinational optimization of the WRF physical
parameterization schemes to improve numerical sea breeze prediction using micro-genetic
algorithm,'' Applied Sciences, vol. 11, no. 23, 11221, 2021.
L. Wang, L. Gu, and Y. Tang, ``Research on alarm reduction of intrusion detection
system based on clustering and whale optimization algorithm,'' Applied Sciences, vol.
11, no. 23, p. 11200, 2021.
C. Hocine and A. Benaissa, ``New binary particle swarm optimization algorithm for
surveillance and camera situation assessments,'' Journal of Electrical Engineering
& Technology, pp. 1–11, prepublish, 2021.
M. Harrabi, O. B. Driss, and K. Ghedira, ``A modified biogeography-based optimization
algorithm with improved mutation operator for job shop scheduling problem with time
lags,'' Logic Journal of the IGPL, vol. 29, no. 6, pp. 951–962, 2021.
M. Li, G. Xu, Q. Lai, and J. Chen, ``A chaotic strategy-based quadratic opposition-based
learning adaptive variable-speed whale optimization algorithm,'' Mathematics and Computers
in Simulation, vol. 194, pp. 71–99, 2022.
F. K. Onay and S. B. Aydemir, ``Chaotic hunger games search optimization algorithm
for global optimization and engineering problems,'' Mathematics and Computers in Simulation,
vol. 194, pp. 514–536, 2022.
F. A. Hashim, E. H. Houssein, K. Hussain, M. S. Mabrouk, and W. Al-Atabany, ``Honey
badger algorithm: New metaheuristic algorithm for solving optimization problems,''
Mathematics and Computers in Simulation, vol. 192, pp. 84–110, 2022.
M. Issa and A. Samn, "Passive vehicle suspension system optimization using Harris
hawk optimization algorithm,'' Mathematics and Computers in Simulation, vol. 192,
pp. 328–345, 2022.
M. Shirzadeh, A. Amirkhani, N. Tork, and H. Taghavifar, ``Trajectory tracking of a
quadrotor using a robust adaptive type-2 fuzzy neural controller optimized by cuckoo
algorithm,'' ISA Transactions, vol. 114, pp. 171–190, 2021.
H. Delavari and S. Naderian, ``Design and HIL implementation of a new robust fractional
sliding mode control of microgrids,'' IET Generation, Transmission & Distribution,
vol. 14, no. 26, pp. 6690–6702, 2020.
M. Jasim and T. Brindha, ``Spinal cord segmentation and injury detection using a crow
search-rider optimization algorithm,'' Biomedical Engineering / Biomedizinische Technik,
vol. 65, no. 3, pp. 293–304, 2020.
L. Abualigah, A. H. Gandomi, M. Abd Elaziz, A. G. Hussien, A. M. Khasawneh, M. Alshinwan,
and E. H. Houssein, ``Nature-inspired optimization algorithms for text document clustering
- A comprehensive analysis,'' Algorithms, vol. 13, no. 12, 335, 2020.
F. Chen, F. Chen, F. Liao, L. Zhang, and J. Zhang, ``Dynamic governance of social
network based on dynamic optimization algorithm: A new perspective of AI system,''
International Journal of Technology Management, vol. 84, no. 1–2, pp. 25–49, 2020.
A. López Martínez and F. J. Cuevas, ``Multiple view relations using the teaching and
learning-based optimization algorithm,'' Computers, vol. 9, no. 4, 101, 2020.
G. B. Rajendran, U. M. Kumarasamy, C. Zarro, P. C. B. Divakara Chari, and S. L. Ullo,
``Land-use and land-cover classification using a human group-based particle swarm
optimization algorithm with an LSTM classifier on hybrid data,'' Remote Sensing, vol.
12, no. 24, 4053, 2020.
J. Karlušić, M. Cipek, D. Pavković, Ž. Šitum, J. Benić, and M. Šušnjar, ``Benefit
assessment of skidder powertrain hybridization utilizing a novel cascade optimization
algorithm,'' Sustainability, vol. 12, no. 24, p. 10396, 2020.
V. M. Karthikeyan, T. Samraj, V. Kandasamy, and N. Chilamkurti, ``Improving the lifetime
of an out-patient implanted medical device using a novel flower pollination-based
optimization algorithm in WBAN systems,'' Mathematics, vol. 8, no. 11, p. 1989, 2020.
M. Rawa, A. Abusorrah, Y. Al-Turki, S. Mekhilef, M. H. Mostafa, Z. M. Ali, and S.
H. E. A. Abdel Aleem, ``Optimal allocation and economic analysis of battery energy
storage systems: self-consumption rate and hosting capacity enhancement for microgrids
with high renewable penetration,'' Sustainability, vol. 12, no. 23, art. 10144, 2020.
N. Fiorentini, M. Maboudi, P. Leandri, M. Losa, and M. Gerke, ``Surface motion prediction
and mapping for road infrastructures management by PS-InSAR measurements and machine
learning algorithms,'' Remote Sensing, vol. 12, no. 23, art. 3976, 2020.
X. Chen and X. Li, ``Fault diagnosis of high-power grid wind turbine based on particle
swarm optimization BP neural network during COVID-19 epidemic period,'' Journal of
Intelligent & Fuzzy Systems, vol. 39, no. 6, pp. 9027–9035, 2020.
Author
He Li was born in 1983, graduated with a bachelor's degree in Education from Sichuan
Normal University, and a master's degree in Linguistics and Applied Linguistics. She
teaches as a lecturer at Chengdu Industry and Trade Vocational and Technical College.
Her main research focuses on early childhood care and education services and management.