Classification and topic tracking of college students’ cybersecurity education based on the internet

Abstract

Campus security issues have always been a hot topic of concern from all walks of life. To analyze the changes in network public opinion of college students’ campus security incidents, this study proposes to classify campus security incidents and track related topics. The classification algorithm and topic tracking algorithm for campus security incidents based on Internet technology are studied, and a prototype system of campus security incidents based on Internet is designed for analysis. The research results show that the Internet-based campus security incident prototype system can classify campus security incidents more accurately, and can also track security incidents on related topics, which provide support for student security management to a certain extent. This technology will be applied to the tracking and classification of college safety accidents, and provide certain reference for ensuring college students’ campus safety.

Keywords

Topic extraction text classification topic tracking security incidents campus security classification prototype system

1. Introduction

Every time a safety accident occurs, it will attract the attention of all sectors of the society, especially the safety accidents that occur on campus or on students [1, 2, 3]. Parents, teachers, social media and many other parties express their own comments and opinions on events thatrelated to their own interests. These views and articles spread all over the world along with the Internet, and eventually forms a complex, influential and uncontrollable social public opinion. However, the network public opinion without specific classification and fixed theme cannot judge the authenticity [4, 5, 6]. To reduce public opinion on campus safety and strengthen the management of student safety work, this research proposes the classification of security incidents based on Internet technology and tracks the subjects of security incidents. The research innovatively combines the text classification topic extraction and topic tracking technology to construct the algorithm design, and uses the prototype system to search and classify the extracted topic words for campus security events. It is expected that this study can provide some reference for the public opinion management of campus security incidents.

2. Literature review

College students’ campus safety is the premise and foundation for the smooth development of various tasks in colleges and universities. Network security in the information age is a huge challenge to campus security. Cohen et al. [7] studied the spread of the new type of coronavirus on university campuses. The results showed that without external intervention, college students would choose to relax in bars or clubs on the basis of their own protection. Daigle et al. [8] investigated the factors of campus crime in the United States and Canada. The results showed that Canadian college students generally believed that their safety was higher than that of American college students, although they faced higher risk of victimization. In addition, the crime factors between the United States and Canada also showed some similarities. Jouriles et al. [9] studied the relationship between the use of party security strategies and the frequency of freshman girls attending parties and being sexually assaulted. The results showed that the frequency of partying was positively correlated with sexual abuse. This relationship was appropriately adjusted through the use of security policies. Murray et al. [10] conducted a study on the prevalence of food insecurity in Australian university campuses and their satisfaction with food choices on campus. The result showed that nearly half of the students were not satisfied with the school’s food supply. The insecurity and scarcity of food environment in university campus were very common.

Chen [11] studied the prevention mechanism of campus security incidents represented by campus bullying incidents from a judicial perspective, aiming at how to reduce the occurrence of campus security incidents. The survey results showed that the construction of the multiple governance system of campus bullying must follow the principles of timely prevention and comprehensive governance. Only by taking measures to deal with the root cause of the incident can effectively avoid the occurrence of relevant safety incidents. Li et al. [12] designed a new sustainable personalized growth system model to manage students’ campus security. The system was based on AI, big data analysis and intelligent graphic element technology to realize intelligent supervision of campus security. The results showed that the method had obvious advantages in accuracy, safety and sustainability. Gong et al. [13] used the feature optimized robot to mine the safety performance of the campus workplace. The results showed that the development of this process feature had significantly improved the safe text mining in the campus workplace and provided support for the campus safety maintenance. Tan et al. [14] used the theory of planned behavior (TPB) to investigate the response of Malaysian college students to potential security threats and prevent major events. The results showed that the TPB method had a good analysis effect and provided a reference for improving students’ ability to deal with unexpected safety events.

Tracking algorithms are popular in engineering, machinery and information classification and extraction because they can achieve accurate positioning of target objects. Khan and Habib [15] used the tracking algorithm to locate the speaker. The results proved that the location tracking algorithm proposed in study can successfully locate the speaker under multiple speakers, and the algorithm had better stability and robustness. Ali et al. [16] used the tracking algorithm to simulate the DC-DC buck converter controlled by dual-loop PI. The results showed that the tracking algorithm had better applicability and more accurate test results when testing the battery charging controller. Srivastava et al. [17] proposed a thunderstorm tracking method based on full flash data and adjacent algorithm to forecast lightning activity in the area of concern. The results showed that the tracking algorithm predicts the area of thunderstorm in advance, with the tracking accuracy of 63%, 80% and 91% respectively for 30 minutes, 15 minutes and 5 minutes. Melzer et al. [18] proposed a recognition tracking algorithm for measuring sea level. The scale vortex tracking algorithm could better explain the embedding environment of the circulation. The results showed that the algorithm could better represent the changes of the upper ocean structure. Liu et al. [19] proposed a hierarchical topic detection and tracking method based on the event itself to solve the problems that were easy to be ignored in the topic detection process and the semantic drift problem, and effectively improved the accuracy of topic detection and tracking.

From the above analysis, the research content on university campus security and target tracking algorithms is rich, but campus security itself contains many potential dangers. The relevant research has not specifically classified and tracked various security events. Therefore, the research combined the topic tracking algorithm classifies campus security incidents and tracks them based on classified subject words, hoping to improve the management mode of campus security incidents and build a better campus environment.

3. Classification of university security events and research on topic tracking algorithms

3.1 Text classification based on thesaurus

The text information classification of campus security incident can be realized by studying the text classification algorithm of subject words. The cosine theorem is used to calculate the similarity of the event texts to be classified in the subject terms of campus security incidents in order to realize the subject vocabulary text classification of campus security incidents. The core of the text classification algorithm based on subject words is to construct the subject words of campus security incidents, establish a vector space model of security events, and then calculate the similarity of the event texts to be classified in the subject terms of campus security incidents. The following describes how thesaurus classification was done for this investigation.

Figure 1.

Workflow chart of text classification based on thesaurus.

Figure 1 is the workflow diagram of text classification of subject words in this research. In practice, it is vital to realize the classification and tracking of campus security incidents because the information about security incidents involving college students is broad and unevenly distributed [20]. First of all, it is necessary to establish a vocabulary related to campus security incident information. In this study, campus security incidents are divided into 12 categories. These categories are education and teaching activities safety, traffic safety, fire safety, food hygiene safety, experimental class safety, school assembly safety, campus violence, gate guard safety, natural disasters, boarding student safety, campus surrounding safety and drowning accidents. Take education and teaching activities as an example. This category includes words such as slapping, corporal punishment and fracture. The traffic safety category includes school bus, rollover and accident.The number of words contained in each category is shown in Table 1. Through the information in Table 1, a more intuitive data reference can be provided for public opinion analysis and processing, providing a reference for establishing a more complete campus security event information text vocabulary. To improve the accuracy of word tagging, the manual tagging method is used for tagging operation, which sacrifices certain time efficiency. Specific information is shown in Table 1.

Table 1

The categories and numbers of campus security incidents

Category	Drowning	Traffic safety	Fire safety	Food hygiene	Experimental safety	Assembly safety
Number	13	147	62	53	41	21
Category	School violence	Doorman security	Natural disaster	Boarding security	Around campus	Teaching activities
Number	89	26	36	35	67	59

Table 1 shows the classification of campus security incidents in this study and the number of words contained in each category. Table 1 illustrates that there are 876 words in the 12 safety incidents, and the words contained in each type of safety incident are different. The most words are used in road safety, while the fewest words are used in drowning occurrences. After completing the establishment of the vocabulary related to campus security incident information, the classification of campus security incidents can begin. The cosine theorem and vector space are the foundations of the categorization technique for incidents involving campus security.

$C=\left\{{t_{1},t_{2},\ldots,t_{i}\ldots t_{n}}\right\}$ is used to represent the vocabulary space vector of campus security incident information, where $t_{i}$ it represents a subject word in the security incident, and the length of the subject word is expressed as $n$ . For each category $C_{j}$ in the table in subject headings, the total number of categories is $m$ . $V(C_{j})=\left\{{u_{1},u_{2},\ldots,u_{i},\ldots,u_{n}}\right\}$ represents the thesaurus vector form, $t_{i}$ represents the weight of $u_{i}$ . When $t_{i}$ belongs to the thesaurus category $C_{j}$ , the value of $u_{i}$ is 1, otherwise it is 0. The weight of each feature item in the thesaurus can be expressed as follows.

$\displaystyle\textit{weight}(t_{\textit{ki}})=2\ast\textit{title\_tf}(t_{% \textit{ki}})+\textit{content\_tf}(t_{\textit{ki}})$ (1)

In Eq. (1), $\textit{content\_tf}(t_{\textit{ki}})$ represents the word frequency of the feature item in this research, and $\textit{title\_tf}(t_{\textit{ki}})$ represents the word frequency of the feature item in the text title. Using the cosine law to calculate the similarity of the event texts to be classified in the security event keywords is as follows.

$\displaystyle\textit{Sim}(D_{i},C_{j})=\frac{w_{1}u_{1}+w_{2}u_{2}+\ldots+w_{k% }u_{k}+\ldots w_{n}u_{n}}{\sqrt{(w_{1}^{2}+w_{2}^{2}+\ldots+w_{k}^{2}+\ldots w% _{n}^{2})\sqrt{(u_{1}^{2}+u_{2}^{2}+\ldots+u_{k}^{2}+\ldots u_{n}^{2}})}}$ (2)

In Eq. (2), the denominator represents the vector product of the feature item set in this study and the vector membrane of each type of topic word, and the numerator represents the membrane product of the feature set and the vector membrane of a certain type of topic word. When Eq. (2) is infinitely close to 1, it means that the subject is closer to the category in the thesaurus.

3.2 Subject extraction based on LDA model

Topic extraction refers to selecting several words in an article to form the main body of the article, so that the main content of the article can be understood through the number of themes in the words [21, 22]. The subject extraction process may be broken down into two primary parts after the aforementioned subject word classification is finished, namely, implementing subject extraction using the LDA model and extracting text events after the subject extraction is finished. When using the LDA model to describe the topic extraction algorithm, D represents all collections of campus security incidents, T represents the collection of security incident topics, and VOC represents the collection of all text in security incidents. In topic extraction, d represents the extracted text, $<w_{1},w_{2},\ldots w_{i},\ldots,w_{n}>$ denotes all the words in the corresponding text, where $w_{i}$ represents the $i$ word in all the words in the text. To calculate the probability distribution of $w_{i}$ under each topic, the d in D is traversed, and the items in d are $w_{i}$ randomly assigned to each topic, and the calculation formula is as follows.

$\displaystyle p_{\textit{wi}}=N_{\textit{wi}}/N$ (3)

In Eq. (3), $p_{\textit{wi}}$ is the $i$ wordprobability in the VOC, $N_{\textit{wi}}$ is the number of the $i$ word in the corresponding topict, and $N$ is the number of all words corresponding topict. Compute and sort $p_{\textit{wi}}$ to find its maximum value. Then calculate the probability values of all $w_{i}$ in the text, and express the probability of all words as $w_{i}=(w_{p1},w_{p2},w_{p3},\ldots,w_{pi},\ldots,w_{pm})$ . A topic is randomly determined in topict, as the initial topic, and the calculation formula of the $i$ probability in $T$ corresponding to $d$ can be expressed as follows.

$\displaystyle p_{ti}=n_{ti}/n$ (4)

In Eq. (4), $p_{ti}$ is the itopic probability in $T$ corresponding $d$ , and $n_{ti}$ is the number in $d$ of the itopic word, $n$ is the total number of all words in $d$ . Then sort the probability sets of each article, and find the maximum value to get the topic of each document. Finally, use $D_{i}=(d_{1},d_{2},d_{3},\ldots,d_{i},\ldots d_{m})$ represents the collection of articles under the corresponding topic, which $d_{i}$ represents the $i$ article in $D$ .

3.3 Subject word extraction based on subject tracking algorithm

When conducting topic tracking, this study uses an improved topic tracking algorithm, which can perform weight calculations based on the characteristics of articles in each category. Match according to the campus security incident vocabulary, and then calculate the security incident words weight in a certain time. After the weights are calculated, the data is sorted by size, and the $N$ subject words of the previous campus safety event are selected to form the campus security vocabulary [23]. The topic tracking algorithm is mainly divided into four parts. Firstly, $D_{i}$ indicates classified articles, $t_{ki}$ refers to the subject words of the corresponding article, which represents the $k$ subject word in the $i$ text, $V(D_{i})=(w_{1i},w_{2i},w_{3i},\ldots w_{ki},\ldots,w_{ni})$ refers to the weight of the classified articles’ subject words. Secondly, compute the word frequency of each topic word in each text and the weight of all topic words by looping through the information for each topic word in the text. The calculation formula is as follows.

$\displaystyle\textit{Sw}(t_{k})=\sum_{i}^{s}{=_{1}w(t_{ki})}$ (5)

In Eq. (5), $t_{k}$ represents the subject words of campus security in the vector space, which $w(t_{ki})$ is the weightof $t_{k}$ in the $i$ chapter. Then the weights obtained are sorted according to the size, and the previous $h$ campus safe words with higher weights are selected. Finally, all the texts are expressed as $D_{i}$ , $t_{i}$ express the previouscampus security words. Each text is traversed to calculate the word frequency $V(D_{i})=(w_{1},w_{2},w_{3},\ldots w_{i},\ldots,w_{n})$ of the security words, and finally the previous $m$ campus security topic word with the largest word frequency can be screened. With the development of events, seed events may produce new states or situations, and the content focus of campus security events has also shifted. This new situation is called a novel event, and call this process the evolution from seed events to novel events. The novel event still belongs to the original theme. When a new event appears in a news report during a real-world news event, there is frequently a retrospective account of the seed event or the preceding new event, and the new event must take place after the seed event. According to this, there is often a retrospective description of the seed event or the previous novel event in the report of the novel event, and the main content of the event description is: time, place, person and other named entities, which means that the named entity vector of the news report has high similarity with the named entity vector of the subject. Set the threshold value as $\lambda_{1}$ , if the similarity of the named entity vectors that are not part of the same topic is lower than the threshold value $\lambda_{1}$ . Then compare the similarity between the content vector of the news report and the content vector of the model. If the similarity is greater than $\lambda_{2}$ , the news report is still considered to belong to the topic.

4. Experimental research and analysis

4.1 Prototype system design of Internet-based network security

To realize the classification and topic tracking of college students’ campus security incidents, this research has designed a prototype system of campus incidents. The system is primarily made up of three modules: a module for thesaurus-based text classification, a module for topic extraction, and a module for topic tracking [24]. In the process of text classification based on thesaurus, there may be two situations. The first is that there is a lack of vocabulary alignment between the subject words and the news, which prevents classification from succeeding, and the second is that there is very little vocabulary alignment between the subject words and the news, which lowers the accuracy of the classification results [25]. To control the classification accuracy, the study uses the following relational expression to express the classification accuracy.

$\displaystyle\textit{precision}\textit{ratio}=\frac{\textit{Number of correct % texts in classification results}}{\textit{Number of texts of classification % reults}}$ (6)

The second module is the topic extraction module. The work of this module is mainly divided into topic event element extraction and topic extraction based on the LDA model. The keywords corresponding to each topic can be extracted through topic extraction based on the LDA model. Taking the traffic safety incident in this research as an example, the number of topics is set to 3. The following table shows the keywords in the traffic safety incident and the number of documents corresponding to each topic.

Table 2

Keywords of traffic safety incidents and the number of corresponding documents

Topic	Key words of traffic safety events	Number of documents
Topic 0	Students, parents, schools, teachers	54
Topic 1	Car, accident, reporter, rescue	39
Topic 2	Accident, police, safety, accident	48

Table 2 shows the keywords of the three topics in traffic safety incidents and the number of documents corresponding to each topic. Through the LDA model algorithm, it is observable that the number of keywords isset first set, the word vector is calculated using TF, and then the keywords can be located. The feature extraction is to divide each topic into multiple subtopics, and finally form a complete event composed of named entities and verbs.

When using the prototype system to analyze network security events, first test the performance of the prototype system used in the study. From Fig. 2, when the threshold value is reduced from 100 mV to 50 mV, the data volume continues to increase until the saturation state. Therefore, the system can complete a data readout at about 6.1 MB/s.

Figure 2.

Performance analysis of prototype system.

Figure 3.

Web page distribution query results.

4.2 Analysis of query results of Internet-based network security prototype system

This research used a prototype system to inquire about college students’ campus security incidents in the online media from May 11 to June 15. The query content includes the media distribution, the proportion of each keyword and the total number of security incidents.

Figure 3 shows the query results and the web page distribution. The statistic shows that the articles about college students’ campus safety events are primarily published in the People’s Daily Online-Education section and were available on this website for a total of 11 days. Compared with this, the articles about university campus safety accidents only appeared for one day on other websites such as Southern Metropolis Daily, The Paper, Dazhong, China News and Huashang Daily. In addition, the study also inquired about the distribution of the number of articles. The results are as follows.

Figure 4.

Distribution of various articles.

Figure 4 shows the distribution of the number of various articles from May 11 to June 15. The figure shows that there are 30 articles on traffic safety accidents, which are the most among all article types. And there are only 5 articles about school violence security incidents. The frequency of the other three types of articles in the figure is relatively close. Through the calculation of the classification accuracy rate, the accuracy rate of the classification results in this study is as high as 80%. In conclusion, the Internet-based network security prototype system suggested in this study may precisely search and categorize campus security incidents involving college students online.

Figure 5.

Tracking results of various types of topic articles.

Figure 5 above shows the tracking results of the top 3 types of thematic articles on security incidents. From the above figure, it can be seen that the top 3 themes of the articles that appeared during the period from May 11 to June 15 are traffic safety, education and teaching activities and security around campus. Among them, Among them, 9 articles on traffic safety appeared in the second week, 10 in the third week, and 9 in the fourth week. On the other hand, there were four educational and teaching theme events in the third week, five in the second week and four in the fourth week. There were two security incidents around campus in the first week and five in the second week.The most articles are related to road safety, and their overall number is the highest. In addition, this study tracked the keywords in the topic articles of traffic safety accidents, and the word frequency search results of the keywords in the prototype system are as follows.

Figure 6.

The occurrence frequency of subject words.

Figure 6 shows the frequency of keywords in the prototype system. Among them, the three keywords “investigation”, “caused accident” and “traffic police” appear most frequently in articles on the topic of traffic safety accidents. The word “survey” appeared continuously during this period, and reached the peak frequency of 9 times in the second week. In four consecutive weeks, the total number of occurrences reached 23. The theme word “accident” appears for three consecutive weeks from the second to the fourth week, of which the number of occurrences of the second week is at most 7, and the total number of occurrences in three weeks is 15. The key word “traffic police” is the same as the word “accident”. It also appeared for three consecutive weeks from the second week to the fourth week, and reached the peak frequency of 8 times in the third week. In the three weeks, it appeared for a total of 15 times.

The analysis results of the four query methods showed that the prototype system designed in this research can accurately classify college students’ campus security incidents and obtain the number of related articles in each category. In addition, the prototype system can also realize the tracking of the subject. After obtaining the relevant subject terms, statistics can be made on these subject terms in various aspects, so that articles related to the target subject can be displayed clearly and intuitively. In short, the LDA model proposed in the study can extract the keywords corresponding to each topic based on the realization of topic extraction. The text event is extracted when the subject extraction is finished, and then a finished event made up of named entities and verbs is created. Finally, the prototype system is used to classify and track topics. The accuracy of the classification results of the prototype system reaches 80%. Therefore, the subject words extracted based on the LDA model can accurately realize the classification and search of college students’ campus security events in the prototype system to meet the needs of campus security management.

5. Conclusion

With the increasing frequency of campus security incidents,all sectors of society pay more attention to student security incidents. A campus security event prototype system based on Internet technology will be developed as part of the research to realize the classification of security events and the tracking of relevant issues in order to better understand how public perception of campus security events develops over time. The feasibility of the system is verified by experiments. The research results showed that the classification accuracy of campus security event prototype system for security events is as high as 80%. In addition, on the basis of ensuring the classification and search of the subject words related to campus security events, the system realizes the tracking of the number and category of security events articles on related topics. This showed that the research and design of campus security event prototype system is feasible in security event classification and subject article tracking, and can provide certain support for the current campus security management of college students. The prototype system can obtain the change trend of college students’ campus security events relatively accurately, provide data support for campus administrators, and is beneficial to the development of campus security education. However, there are still some deficiencies in this study. The subject words used in the study are all collected manually, so it is inevitable to have a certain degree of subjectivity.In the future research, more safety events need to be involved in the collection of safety event keywords to improve the rigor of the research. In addition, when using the prototype system to track the subject words, more subject words need to be added to improve the scope of security events covered by the system.

Footnotes

Funding

This study was funded by the National Social Science Foundation of China (Title: Study on Theoretical Logic and Practical Path of Cultural Mission of Ideological and Political Education in New Era. Grant No. 20XKS022).

References

Blazic

. Cybersecurity skills among European high-school students: A new approach in the design of sustainable educational development in cybersecurity. Sustainability. 2022; 14(8): 4763.

Calitz

Cullen

MDM

Jooste

. The influence of safety and security on students’ choice of university in South Africa. J Stud Int Educ. 2020; 24(2): 269-285.

Lei

Yang

Byrne

Jiang

Liu

Wang

. Family cohesion and social adjustment of Chinese university students: The mediating effects of sense of security and personal relationships. Curr Psychol. 2021; 40(4): 1872-1883.

Birnhack

Perry-Hazan

. School surveillance in context: High school students’ perspectives on CCTV, privacy, and security. Youth Soc. 2020; 52(7): 1312-1330.

Tilley

Kolodetsky

Cottrell

Tilton

. Correlates to increased risk of sexual assault and sexual harassment among LGBT plus university students. J Forensic Nurs. 2020; 16(2): 63-72.

Soldavini

Berner

. P22 Prevalence of and characteristics associated with food security status among international college students. J Nutr Educ Behav. 2020; 52(7): S25-S26.

Cohen

Hoyt

Nichols

Yazdani

Doston

. Opportunities to reduce young adult college students’ COVID-19-related risk behaviors: Insights from a national, longitudinal cohort. J Adolescent Health. 2021; 69(3): 383-389.

Daigle

Hancock

Chafin

Azimi

. US and Canadian college students’ fear of crime: A comparative investigation of fear of crime and its correlates. J Interpers Violence. 2022; 37(15-16): NP12901-NP12932.

Jouriles

Krauss

Sargent

Nguyen

Cascardi

Grych

McDonald

. Party frequency, party-safety strategies, and sexual victimization among first-year female college students. J Am Coll Health. 2022; 70(6): 1788-1793.

10.

Murray

Peterson

Primo

Elliott

Otlowski

Auckland

Kent

. Prevalence of food insecurity and satisfaction with on-campus food choices among Australian university students. Int J Sust Higher Ed. 2021; 22(4): 731-746.

11.

Chen

. Research on the governance of campus bullying and prevention mechanism of students’ mental health problem. J Community Psychol. 2022; 50(8): 3542-3559.

12.

Xue

Tseng

Wang

. A personal growth system supporting the sustainable development of students based on intelligent graph element technology. Sustainability. 2022; 14(12): 7196.

13.

Gong

Samuel

RDJ

Pandian

. Meta-heuristic feature optimization for ontology-based data security in a campus workplace with robotic assistance. Work. 2021; 68(3): 913-922.

14.

Tan

Sia

JKM

Tang

KHD

. Examining students’ behavior towards campus security preparedness exercise: The role of perceived risk within the theory of planned behavior. Curr Psychol. 2022; 41(7): 4358-4367.

15.

Khan

Habib

. Concurrent speakers localization using blind source separation and microphone array geometry. Multidim Syst Signal P. 2021; 32(4): 1159-1184.

16.

Ali

Premkumar

Vishnupriya

Manikandan

Thamizhselvan

. Design and development of realistic PV emulator adaptable to the maximum power point tracking algorithm and battery charging controller. Sol Energy. 2021; 220: 473-490.

17.

Srivastava

Liu

Yuan

Wang

Babalola

, et al. Lightning nowcasting with an algorithm of thunderstorm tracking based on lightning location data over the Beijing area. Adv Atmos Sci. 2022; 39(1): 178-188.

18.

Melzer

Jensen

Rydbeck

. Evolution of the great whirl using an altimetry-based eddy tracking algorithm. Geophys Res Lett. 2019; 46(8): 4378-4385.

19.

Liu

Jiang

Tang

. Topic detection and tracking based on event ontology. IEEE Access. 2020; 8: 98044-98056.

20.

Wang

Liu

Zhang

Feng

. A multi-label text classification method via dynamic semantic representation model and deep neural network. Appl Intell. 2020; 50(8): 2339-2351.

21.

Joshi

Mittal

Joshi

. Improving the performance of semantic graph-based keyword extraction and text summarization using fuzzy relations in Hindi Wordnet. J Intell Fuzzy Syst. 2022; 43(3): 3771-3788.

22.

Bai

. Automatic content extraction and time-aware topic clustering for large-scale social network on cloud platform. J Supercomput. 2019; 75(5): 2890-2924.

23.

Gopalan

Linden-Carmichael

Lanza

. College students’ sense of belonging and mental health amidst the COVID-19 pandemic. J Adolescent Health. 2022; 70(2): 228-233.

24.

Borisov

Braginets

Voronin

Lasitsa

Semenikhin

Fedorov

. A crystal-diffraction system prototype for measuring neutron’s charge: Thermal protection test and vibration measurements. Tech Phys Lett. 2019; 45(1): 53-56.

25.

Monno

Teranaka

Yoshizaki

Tanaka

Okutomi

. Single-sensor RGB-NIR imaging: High-quality system design and prototype implementation. IEEE Sens J. 2019; 19(2): 497-507.