Deep Learning for Sea Turtle Classification: A Bibliometric Analysis Using VOSviewer

Deep learning in the context of sea turtles constitutes a significant area of exploration within marine science. Consequently, the aim of this study is to perform a bibliometric analysis focusing on the subject of sea turtle deep learning, leveraging mapping analysis through the utilization of VOSviewer software. For this research, we employed a bibliometric and descriptive quantitative approach. The data was acquired by conducting a search on Google Scholar using the keyword "Sea Turtle Deep Learning," which yielded a total of 880 articles published between 2018 and 2023. Notably, only 19 of these articles were directly relevant to the research topic. The findings underscore the diversity of research outcomes in the realm of sea turtle deep learning over this time span. In conclusion, this investigation underscores the significance of conducting bibliometric analyses, particularly within the domain of sea turtle deep learning, and serves as a valuable reference for future research endeavours in defining research themes.


INTRODUCTION
Sea turtle research, based on deep learning, has gained significant traction among marine scientists due to the current vulnerable and endangered status of sea turtles, as noted in (Kittinger et al., 2013).Identifying individual sea turtles holds crucial importance for comprehending their behavior, life history, evaluating threats, assessing anthropogenic impacts, and understanding population dynamics (Blumenthal et al., 2009;Wallace et al., 2010;Casale et al., 2013).Deep learning serves as a technique for classifying and identifying sea turtles by employing image recognition methods in computer vision.Previous studies have indicated the widespread use of deep learning in the classification and recognition of sea turtles.Although a limited number of papers have been published on sea turtle identification, focusing primarily on species recognition (Attal & Direkoglu, 2020;Liu et al., 2020;Fauria et al., 2023), one particular study (Papafitsoros et al., 2022) emphasized the detrimental effects of tagging on sea turtles.This study remains the only literature found via manusal internet searches to date.However, there are immense opportunities to explore specific themes and conduct extensive research in this field.Given the importance of the topic, it is crucial to conduct a bibliometric study to examine this subject.Bibliometric studies can track trends, identify influential authors or publications, measure the impact of research, and understand the interconnections between different areas of study (Butticè, & Ughetto, 2023).VOSviewer (Arruda et al., 2022) (Coulibaly et al., 2022).In the healthcare field, Khairi et al. proposed Deep Learning on Histopathology Images for Breast Cancer Classification (Khairi et al., 2021), where they focused on annual publication trends, co-authorship networks among countries, authors, and scientific journals.They analyzed the cooccurrence network of authors' keywords for potential future directions in the field.Nevertheless, research on bibliometric analysis concerning sea turtle classification remains scarce, especially regarding the utilization of VOSviewer software as a tool for conducting mapping analysis.Therefore, this research aims to conduct a bibliometric analysis in the field of sea turtle classification, integrating mapping with VOSviewer software.This analysis is crucial to ascertain the quantity and novelty of data and is expected to serve as a valuable resource for academics in determining and selecting research subjects, particularly those related to sea turtle classification.

MATERIALS AND METHOD
The research employed bibliometric and descriptive quantitative approaches.Data were collected from published journals indexed by Google Scholar due to its convenient accessibility.VOSviewer 1.6.17 was utilized to gather the data, filter relevant materials, and extract information specifically related to deep learning sea turtle classification.By inputting the keyword "deep learning sea turtle" in VOSviewer, we retrieved 880 data entries, which were subsequently refined to 19 related journals.The study encompassed publications from 2018 to 2023, with the collected articles stored in *.ris format.Subsequently, we employed VOSviewer software to visualize and analyze bibliometric trends through the generated bibliometric maps.The article data from the compiled database sources were mapped using the VOSviewer software, which facilitated the categorization of data visualization into network, overlay, and density formats.Furthermore, the terms included in the VOSviewer mapping visualization were filtered to ensure relevance and accuracy.Specifically, the number of articles in 2018 was 112, which then increased to 160 in 2019.The upward trend continued with a further increase to 236 articles in 2020.Subsequently, in 2021, the number of articles continued to rise to 241.However, there was a decrease in 2022, with the number of articles dropping to 158.The trend continued in 2023, with the number of articles plummeting to 52.

Visualization of deep learning sea turtle topic area using VOSviewer
According to Al Husaeni and Nandiyanto (Husaeni & Nandiyanto, 2023), in the VOSviewer application, the minimum number of relationships is set by 2 terms.However, in this study, the minimum number of relationships in the VOSviewer between terms is 3. Therefore, the results obtained are 19 items with a total of 5 clusters.Research related to deep learning sea turtle based on visualization mapping analysis is divided into 5 clusters, namely:  6).
Cluster 1 is displayed in red, Cluster 2 is displayed in green, Cluster 3 is displayed in blue, Cluster 4 is displayed in yellow, and Cluster 5 is displayed in purple.

Network visualization deep learning sea turtle topic area using VOSviewer
Within the VOSviewer application, term mapping is categorized into three types, one of which is Network Visualization.This form of visualization illustrates the interrelationships among terms on a map.Connections between terms are represented by lines or networks extending from one term to another.Figure 8 presents the Network Visualization of the term 'Deep Learning Sea Turtle' derived from the VOSviewer application.Additionally, Figure 7 depicts the visualization of each cluster within the researched topic areas.Notably, as observed in Figure 7, the term 'deep neural network' exhibits the highest connectivity, being prevalent across all clusters.

Overlay visualization deep learning sea turtle topic area using VOSviewer
Apart from Network Visualization, the VOSviewer application also offers an overlay depiction as a form of mapping.Overlay Visualization focuses on assessing the novelty of a term within the research context.Figure 8 demonstrates the representation of term novelty in research related to 'Deep Learning Sea Turtle'.In this type of term mapping, the popularity of a term over the years is illustrated.The colors in the Overlay Visualization represent the recency of a term within a specific timeframe.Within the scope of this study, the period from 2018 to 2023 was considered.A darker color tending towards purple signifies that research on a term was more prevalent closer to 2018.On the other hand, a lighter color leaning towards yellow indicates that the term appeared more frequently in recent research.The overlay visualization effectively demonstrates the evolution of studies over time and the interrelationships among these terms.

Desity visualization of Deep Learning Sea Turtle
The VOSviewer application encompasses the last type of mapping depiction known as Density Visualization.
This representation categorizes each term based on its prominence within research.Figure 9 illustrates the Density Visualization of 'Deep Learning Sea Turtle'.This form of mapping is discernible through the color gradations of each term.A lighter color signifies that research on a particular term is gaining prominence or increasing.Conversely, a darker or more faded color indicates a decline or reduced frequency in research related to that term.In Figure 9, several terms, including 'deep neural network', 'detection', and 'neural network', appear as yellow with significant diameters.This suggests that these terms are frequently employed within existing studies.The density map, derived from the analysis of all articles on deep learning sea turtles from 2018 to 2023, reveals a pattern of yellow shades.The intensity of the yellow color and the size of the circle correlate with the frequency of these keywords.Conversely, fading or blending colors against the green background indicate that these keywords appear less frequently.This outcome underscores the efficacy of bibliometric analysis in exploring and visualizing the current literature, thereby providing a foundation for deciding the need for further research.

Figure 1
Figure 1 illustrates the progression of deep learning sea turtle research from 2018 to 2023.As indicated in the figure, the development of deep learning sea turtle research over the past five years experienced growth from 2018 to 2021, followed by a subsequent decline in 2022.Notably, in 2023, there was a significant decrease in the number of publications.

Fig. 9 .
Fig. 9. Density visualization of deep learning sea turtle.4. CONCLUSION This study aims to conduct a bibliometric analysis of the literature concerning deep learning sea turtles.The keyword "deep learning sea turtle" was employed to retrieve data, encompassing relevant topic areas within abstracts, titles, and keywords.After data filtration, a selection of 19 pertinent articles was obtained.The mapping data was processed the VOSviewer software, generating network, overlay, and density visualizations.Analysis of the mapping findings through VOSviewer indicated a declining trend in sea turtle research involving deep learning from 2018 to 2023.This study employed bibliometric methods to identify key themes within previous research, offering valuable insights for evaluating the potential areas of novelty in future research.

Table 1 .
The table displays data pertaining to the selected 19 articles, revealing that the maximum citations for a single article in the realm of deep learning sea turtles reached 1974 in 2018, while the lowest was recorded at 168 in 2020.Table1shows summarized of all the 19 articles.