Human-centric Computing and Information Sciences volume 12, Article number: 59 (2022)
Cite this article 1 Accesses
https://doi.org/10.22967/HCIS.2022.12.059
Deep learning technologies for mobile networks have been regarded as a popular issue due to the development of mobile technology, deep learning, and mobile smart applications. This paper investigates research trends in deep learning technologies for mobile networks in computer science and technology research by analyzing the articles published from 2016 to 2022 in the journals included in the Web of Science (WoS) database. That is, SSCI/SCI journal articles were collected to analyze the most frequently cited relevant articles, countries, authors, institutions, and application areas, and to investigate the international research trends of deep learning technologies for mobile networks via VOSviewer. The analysis results identified four main collaborating institutions, namely King Saud University, Beijing University of Post & Telecommunications, the University of Electronic Science & Technology of China, and Xidian University, as well as four keyword clusters, covering the Internet of Things, edge computing, and deep learning. The results also show that the study of Li, Ota, and Dong's (2018) is the most frequently cited article in the field of deep learning technologies for mobile networks. The main application areas are Engineering Electrical Electronic, followed by Computer Science Information Systems and Telecommunications. The results of this study highlight the international research trend of deep learning technologies for mobile networks, and provide new insights for researchers.
Deep Learning, Mobile Network, Cloud Computing, Edge Computing, Internet of Things, IoT, Technology
The rapid development of information technology (IT) has a critical impact on the Internet of Things (IoT) [1]. With the advancement of mobile technology, deep learning has been widely applied in computer vision, natural language processing, IoT edge computing, and big data analytics. Thus, deep learning is recognized by scholars as a superior method to recognize and detect targets compared to traditional methods such as smart phones and IoT sensors [2]. Deep learning methods require inferencing and training processes, as well as a large number of resources to calculate edge computation in real time. Gan et al. [3] provided an overview of the latest techniques about deep learning and edge computing, applications of deep learning in networks, various methods for fast execution of deep learning through terminals, a combination of deep learning inference servers at the edge, and the cloud. They also described deep learning models and training methods across multiple edge devices. In addition, they identified challenges in network technology and management, benchmarking and privacy, and system performance. Also, there will be challenges in network techniques and management, benchmarking and privacy, as well as system performance. Chen and Ran [4] enable research scholars to realize the usefulness of deep learning at the edge of the network, learn about the most commonly used techniques for accelerating deep learning inferences and execution for distributed training on edge devices, and grasp trends and opportunities. For example, artificial intelligence and Internet of Things (AI-IoT) applications and innovations have had the greatest impact on healthcare, affecting the way patients interact with healthcare services during the coronavirus disease 2019 (COVID-19) epidemic [5]. Many researchers working on COVID-19 have aimed to combine mobile networks with edge computing technology to meet the needs of market consumers and improve the IoT in the global market [6].
In this paper, the Social Science Citation Index (SSCI) journals from 2016–2022 in the Web of Science (WoS) database were searched with bibliometric analysis to identify literature on the research areas of deep learning, mobile learning, and edge computing at the same time [7]. In 2016, Sheng et al. [8] proposed the k-degree layer-wise network for geographically distributed computing between the cloud and the IoT; their article was published in the international journal, IEICE Transactions on Communications. Since then, research on deep learning and computing between the cloud and the IoT has become an attractive field for researchers. In a recent review on the algorithms in deep learning with edge computing, Dargazany et al. [9] comprehensively reviewed wearable IoT-related research on deep learning for big data analytics. Chen and Ran [4] pointed out the current development of deep learning at the edge with methods for fast inferencing. Atitallah et al. [10] systematically reviewed the support of deep learning and IoT big data analytics for smart city development to explore the development of IoT big data in the past few years. Based on previous studies, such review articles have been limited to specific areas (e.g., healthcare systems, smart cities, or specific technologies), but have not performed bibliometric analysis of the literature.
In addition, there are limitations to these review studies, most of which did not adopt either a clear search or analysis process. These studies also did not clearly analyze articles from non-international journals [9, 11]. Thus, there is a need to adopt bibliometric analysis to address the limitations of review articles in the field of information engineering, and to analyze the unique contributions and trends of research scholars in the field.
As is known from the literature, bibliometric analysis is considered as an effective analytical technique to assess the achievements of a specific academic research area and to highlight the unique value and contributions of research scholars in various fields [12]. Meanwhile, the results of the quantitative analysis of the literature can easily provide future scholars with a clear understanding of previous articles and serve as a prediction of research trends in specific areas. Nowadays, bibliometric analysis is widely used in cross-field scientific research and trend analysis of emerging topics for specific studies [13]. For example, Su et al.
[14] applied bibliometric analysis of core competencies associated with nursing management publications to summarize the core competencies required for international nursing; their article was published in the International Journal of Nursing Management. Hwang and Chen [15] reviewed seven SSCI educational journals from 1990 to 2019 via bibliometric and visual analysis to quantitatively review game-based learning, and illustrated the clustering and trends of gaming and educational technologies. According to the InCites JCR Report, bibliometric analysis is an effective method to elucidate research trends, and can provide researchers with suggestions for future research on deep learning technologies for mobile networks (DLMNs). Therefore, it is worthwhile to explore the research trends and technology development of DLMNs.
For the past 5 years, deep learning has been used as a fundamental technique for AI-IoT technologies by academics. Through deep learning, it is possible to separate the privacy information mixed in with the raw data and analyze health-related data, as well as privacy preservation and data analytics in IoT-enabled healthcare systems [16]. Since the global IoT has risen, the new intelligent integration between an IoT platform and deep learning neural network (DNN) algorithm for the online monitoring of computer numerical control (CNC) machines allows IT personnel to conduct goal-oriented data analysis through emerging platforms [16]. Several studies have indicated that deep learning is an effective method for analyzing various types of data, allowing researchers to obtain meaningful content within a platform [17]. Based on application results for mobile edge computing, some studies have shown the impact of deep learning on the performance of collaborative high-speed reasoning, including cloud computing, edge computing, and the IoT [18–20]. Xu et al. [18] proposed an offloading framework for deep learning edge services in 5G networks, and defined the formalized concepts in order to build a computational migration optimization model. Fagbohungbe et al. [19] proposed a novel privacy-preserving edge intelligent computing framework for image classification in the IoT. These results further showed that deep learning is a useful technique for image classification. For example, Singh et al. [21] applied AI-based mobile edge computing for the IoT. Thus, mobile edge computing makes the system more robust, flexible, efficient, and accurate. Deep learning has been proven to train different models for mobile edge computing. However, mobile edge computing also faces several challenges, including security and privacy, deployment protocols, and offload management. Moreover, most researchers miss combining deep learning with interesting aspects of mobile edge computing, as their work is usually limited to a single aspect. Thus, a comprehensive literature review is needed on the aspects of deep learning via mobile edge computing.
On the basis of previous studies, less bibliometric analysis has been conducted to investigate the research trends of DLMNs. As mentioned above, if we can understand the application trends of deep learning technologies for mobile networks, it will be easier for researchers to understand the cross-disciplinary international research directions, research interests, and the evolution of information technology. Therefore, this study applied bibliometric analysis of the trends in published papers in SSCI/SCI international journals on deep learning and mobile networks in the field of computer science and technology from 2016 to 2022 to answer the following research questions:
(1) What are the top journals publishing in the DLMNs research?
(2) What are the top 10 most frequently cited articles in the DLMNs research?
(3) What are the top 10 publishing countries in the DLMNs research?
(4) Who are the most published authors in the DLMNs research?
(5) Which institutions and fields have made the greatest contribution to the DLMNs research?
(6) Which keywords form the main clusters in the DLMNs research?
Search Criteria
This study searched the WoS database with the keywords "deep learning" and "mobile network" from 1987 to 2022, and obtained relevant journal publications starting in 2016. The search criteria were set as "deep learning" and "mobile network" OR "Cloud computing" OR "Edge computing" and "Internet of Things" OR "IoTs". The data collection flow of the study is depicted in Fig. 1. Fig. 1 shows the article data collection process. Studies published in SSCI/SCI journals were included, and a total of 476 articles were found between 2016 and 2022.
Data Extraction and Analysis
In this paper, 476 publications were obtained from the WoS database and were imported into the VOSviewer software for constructing and visualizing bibliometric networks. Through the visualized graphics, researchers can quickly identify the files in the database and the meaningful graphs [22] to perform, for example, author, institution, country, and keyword co-occurrence analysis. Researchers have also mentioned that citation analysis can contribute as a pilot study to identify specific professional fields [23]. According to the previous references of bibliometrics analysis, if published articles are frequently cited, it is reasonable to infer that the authors and research directions of these articles are important indicators of the development of the specific professional fields [24]. In addition, research scholars can clearly understand the trends and historical background of the specific field of DLMNs through the citation rate of published articles and the frequent occurrence of keywords, and inspire new researchers to trace and explore the potential research issues of DLMNs research in the IT research field [25].
Data Distribution
Rank | Publication source | Authors | Title | Technical environment | Number of citations |
1 | IEEE Network | Li et al. [27] | Learning IoT in edge: deep learning for the Internet of Things with edge computing | IoT | 716 |
2 | Proceedings of the IEEE | Zhou et al. [28] | Edge intelligence: paving the last mile of artificial intelligence with edge computing | Edge computing | 506 |
3 | IEEE Communications Surveys and Tutorials | Lim et al. [29] | Federated learning in mobile edge networks: a comprehensive survey | MEN | 376 |
4 | IEEE Communications Surveys and Tutorials | Wang et al. [30] | Convergence of edge computing and deep learning: a comprehensive survey | Edge computing | 297 |
5 | Future Generation Computer Systems: The International Journal of eScience | Tuli et al. [31] | HealthFog: an ensemble deep learning based smart healthcare system for automatic diagnosis of heart diseases in integrated IoT and fog computing environments | IoT & fog computing | 159 |
6 | IEEE Internet of Things Journal | Deng et al. [32] | Edge intelligence: the confluence of edge computing and artificial intelligence | Edge computing | 153 |
7 | IEEE Transactions on Wireless Communications | Li et al. [33] | Edge AI: on-demand accelerating deep neural network inference via edge computing | Edge computing | 149 |
8 | IEEE Internet of Things Journal | Zhang et al. [34] | Deep learning empowered task offloading for mobile edge computing in urban informatics | Edge computing | 144 |
9 | IEEE Communications Magazine | Abeshu and Chilamkurti [35] | Deep learning: the frontier for distributed attack detection in fog-to-things computing | Fog computing | 141 |
10 | IEEE Internet of Things Journal | Zhou et al. [36] | Deep-learning-enhanced human activity recognition for Internet of healthcare Things | IoT | 124 |
Cluster 1 | Cluster 2 | Cluster 3 | Cluster 4 |
Internet of Things Cloud computing Internet Things Internet of Things (IoT) Big data Fog computing Optimization Model Sensors |
Edge computing Computational modeling Training Servers Task analysis Computer architecture Data models Cloud Image edge detection |
IoT Networks Security System Intrusion detection Privacy Blockchain Framework |
Deep learning Machine learning Artificial intelligence Feature extraction Neural networks |
In the last 5 years, information development and resources have become more diverse due to the advancement of IT, the IoT, blockchain, and globalization in the new technological era. In particular, DLMNs with the IoT have provided innovative technologies [38]. This study analyzed the most cited international journal articles on deep learning in mobile networks in the WoS database, allowing researchers to identify critical articles, journals, authors, countries, institutions, keywords, and domains. In this paper, a bibliographic and visual analysis was conducted to identify the connection between deep learning for mobile network research and new information technologies, which could help scholars solidify their knowledge of the research trends in this research domain.
As shown in Fig. 2, the overall trend of DLMNs with the IoT has grown significantly since 2019. Regarding international journals, the IEEE Internet of Things Journal and IEEE Access are the top journals in the field of computer engineering research, followed by IEEE Transactions on Industrial Informatics and Electronics. Meanwhile these four SSCI/SCI journals are frequently cited due to their high-quality research with IFs of 10.238, 3.476, 11.648, and 2.69, respectively for the last 5 years. It implies that researchers are struggling in deep learning in mobile network research published in these international core journals [32, 34, 36].
In terms of the most frequently cited articles related to DLMNs with the IoT, this study found that Li et al.’s research [27] had the highest citation rate in 2018 (716 citations). Li et al.’s research presented a powerful technical tool for deep learning for the IoT with edge computing. From the analysis of Li et al.’s study, deep learning can be used as an edge computing tool to improve the effectiveness of the IoT and can influence the development of computer engineering. Following their research, Zhou et al. [28] (506 citations) proposed the concept of edge intelligence by combining AI with edge computing, which may be the reason for the large number of citations. Table 1 lists the top 10 most cited articles related to DLMNs. This finding is consistent with the findings of Chang and Chu [13], indicating that researchers can follow high-citation studies to identify research trends and issues and to inspire their own research.
As for the most published articles according to the first author’s country, as shown in Fig. 5, the top three countries in terms of most publications are the People’s Republic of China, followed by India, the USA, Saudi Arabia, and South Korea. Authors who published the most are C. S. Hong, who published three, three, and two papers in 2020, 2021, and 2022, respectively. For example, Suhail et al. [39] applied trustworthy digital twins in the industrial IoT with blockchain. Munir et al. [40] pointed out that the IoT in blockchain has strong potential to improve system performance.
As for the partner institutions, Fig. 6 shows the visualized results of affiliation occurrences. King Saud University, Beijing University of Post & Telecommunications, the University of Electronic Science & Technology of China, and Xidian University are the four major partner institutions. For example, Prince Sattam Bin Abdulaziz University is a collaborative institution led by King Saud University [41] Their cross-institutional collaboration uses deep learning-based data offloading and network attack detection techniques to enhance the quality of experience for mobile edge computing. They proposed a comprehensive set of simulations and the resultant experimental values highlighted the improved performance of deep learning. Therefore, it should be considered how institutions or multinational collaborations can break through language barriers or information technologies to design experiments in the future. In addition, Fig. 8 shows that Engineering Electrical Electronic, Computer Science Information Systems, and Telecommunications are the most common application fields. For example, Wang and Tong [42] used deep learning and the internet of things to analyze students' high-level dance movement performance in a series of dance activities. It can be argued that deep learning combined with the IoT is worth being promoted in various fields, such as specific fields like Food Science Technology, Green Sustainable Science Technology, Surgery, or Thermodynamics.
The most frequently occurring keywords help researchers to find relevant research topics in databases. From the analysis of this study, it was found that the main author keywords were divided into four main areas: Internet of Things, edge computing, IoTs, and deep learning. For example, in terms of "Internet of Things" and "edge computing," Vijayasekaran and Duraipandian [1] effectively improved the clustering and deep learning-based resource scheduling for edge computing to integrate cloud-IoTs. Therefore, the most frequently used keywords help researchers explore the research trends and find out the most popular themes. In addition, Alam et al. [43] illustrated the impact of video big data analytics in the cloud. Their study received the highest number of citations for DLMNs with the IoT in the computer science research. Therefore, the results of this study can provide specific information for subsequent researchers to understand the global impact of DLMNs with the IoT research.
Based on the analysis results, DLMNs with the IoT is a powerful information technology that enhances the value of data analysis through computer vision, natural language processing, IoT terminals, and big data. In addition, the international journal analysis proves that the IEEE Internet of Things Journal is not only an internationally recognized journal, but also has a recognized influential position in computer technology research for Q1, which helps lead the global trend. As the above findings show, the research on DLMNs with the IoT deserves further analysis and exploration, and future research can investigate how IT and researchers can combine various cloud systems or models to analyze useful data to enhance information security and assist in decision making.
Strengths and Limitations
The main contribution of this study is that it integrated bibliographic analysis with the VOSviewer software to explore DLMNs with the IoT in the field of computer technology. This study is the first research to explore DLMNs with the IoT using a bibliographic analysis in the last 5 years. There are, however, still some research limitations to this bibliographic analysis. First, the study obtained SSCI/SCI journals as samples only from the WoS database. The analysis results may be different if other databases can be included with the same analysis method. Furthermore, the findings of the study revealed no article related to DLMNs with the IoT research before 2016. Until 2019, Sheng et al. [8] published a related article, and the number of publications showed an increase from 2019. In addition, this study only used the keywords "deep learning" and "mobile network" OR "Cloud computing" OR "Edge computing" and "Internet of Things" OR "IoT," which may have limited the focus to Engineering Electrical Electronic, Computer Science Information Systems and Telecommunications. New research trends may emerge if all articles in the database can be searched for keywords that have been used before.
From the analysis results of this research, DLMNs with the IoT proved to be advantageous for various domains, and this paper reports the important articles, journals, authors, countries, institutions, keywords, and topics covered by the domain. In recent years, the use of bibliographic analysis to explore trends in a particular area of expertise has received increasing attention, including investigating the general trends of publications in specific area, prolific authors and countries, most cited articles, and keywords occurrences, etc. [44]. For example, Zhang and Aslan [45] reported the state of artificial intelligence in education (AIEd) research published in 1993–2020. Wang et al. [25] investigated the 100 most cited chatbot-related human behavior research. Moreover, several researchers, such as Hwang et al. [46] and Chiu et al. [47] have indicated the importance and potential of conducting emerging technology related studies. They figured out that, in the digital age, cultivating information literacy, and cross-team cooperation education is fundamental. They also suggested that information and R&D personnel should combine and discuss together to develop from the perspective of user customization, collect and develop emerging valuable issues, and revert to other organizations or cross-country collaboration to implement DLMNs with the IoT technology in a cross-country and multi-cultural context, to enhance information literacy and work for the global welfare.
Conceptualization, HCChu, CYC, HCChao. Methodology, HCChu. Software, validation, and formal analysis, CYC, HCChu. Investigation, CYC, HCChu. Resources, CYC, HCChu. Data curation, CYC, HCChu. Writing—original draft preparation, CYC, HCChu. Writing—review and editing, CYC, HCChu. Visualization, CYC, HCChu. Supervision, HCChao. Project administration, HCChao.
This research was funded by the Ministry of Science and Technology of Taiwan (No. Most-111-2410-H-038-029-MY2, MOST-109-2511-H-011-002-MY3, and MOST-111-2221-E-259-007).
The authors declare that they have no competing interests.
Name : Hui-Chun Chu
Affiliation : Soochow University, Taiwan
Biography : Hui-Chun Chu is currently a Distinguished Professor at the Department of Computer Science and Information Management in Soochow University, where she also serves as the chair. Her research interests include mobile and ubiquitous learning, game-based learning, flipped learning, technology-assisted health care education, AI in medical diagnosis and knowledge engineering in education. She also served as the Associate Editor of IEEE Transactions on Learning Technologies (SSCI, Q1) since 2015 and served as the guest editor of Interactive Learning Environments (SSCI, Q1) in 2013 and 2016. Moreover, Dr. Chu received the reward of “The top 50 Flipped Learning leaders in higher education worldwide” in 2018.
Name : Ching-Yi Chang
Affiliation : Taipei Medical University, Taiwanbr/>
Biography : Ching-Yi Chang is currently an assistant professor at the Department of School of Nursing, College of Nursing, Taipei Medical University. Dr. Chang serves as an editorial board member and a reviewer for more than 29 educational technology journals. Her research interests include maternal and child health, mobile learning, digital game-based learning, flipped classrooms, medical education, nursing education, and AI in education. Dr. Chang has published more than 32 academic papers. Among those publications, more than 28 papers are published in SSCI/SCI journals. Owing to her reputation in academic research and innovative inventions in education, she received the research award from the Taipei Medical University of Taiwan in the year 2021.
Name : Han-Chieh Chao
Affiliation : National Dong Hwa University, Taiwan
Biography : Han-Chieh Chao received his M.S. and Ph.D. degrees in Electrical Engineering from Purdue University, West Lafayette, Indiana, in 1989 and 1993, respectively. He is currently a professor with the Department of Electrical Engineering, National Dong Hwa University, where he also serves as the president. He was the Director of the Computer Center for Ministry of Education Taiwan from September 2008 to July 2010. His research interests include IPv6, Cross-Layer Design, Cloud Computing, IoT, and 5G Mobile Networks. He has authored or co-authored 4 books and has published about 400 refereed professional research papers. He has completed more than 150 MSEE thesis students and 11 Ph.D. students. He served as the Editor-in-Chief for the Institution of Engineering and Technology Networks, the International Journal of Internet Protocol Technology, and the International Journal of Ad Hoc and Ubiquitous Computing. He is a Fellow of IET (IEE) and a Chartered Fellow of the British Computer Society. Dr. Chao has been ranked as the top 10 Computer Scientists in Taiwan for 2020 by Guide2Research. Due to Dr. Chao’s contribution of suburban ICT education, he has been awarded the US President's Lifetime Achievement Award and International Albert Schweitzer Foundation Human Contribution Award in 2016, and South East Asia Regional Computer Confederation, SEARCC in 2017.
Hui-Chun Chu1, Ching-Yi Chang2,3, and Han-Chieh Chao4,*, Mapping Deep Learning Technologies for Mobile Networks with the Internet of Things, Article number: 12:59 (2022) Cite this article 1 Accesses
Download citationAnyone you share the following link with will be able to read this content:
Provided by the Springer Nature SharedIt content-sharing initiative