TL;DR
This paper provides a comprehensive visualization of the big data research landscape from 2002-2016 using bibliometric analysis. It maps the intellectual structure, identifies key players (journals, institutions, authors), and highlights emerging research hotspots, offering a roadmap for future sustainable development in the field.
What Problem Does This Paper Address?
The explosion of big data research, fueled by advancements in IoT, social networks, and cloud computing, has led to a massive increase in publications. This rapid growth makes it challenging for researchers, policymakers, and industry professionals to grasp the current state of the field, identify emerging trends, and understand the intellectual underpinnings of big data research. Without a clear overview, it's difficult to pinpoint research gaps and direct efforts towards sustainable development. This paper aims to solve this by providing a visual map of the big data research ecosystem.
Methods
The authors employed a bibliometric approach, a quantitative analysis of scholarly literature. They utilized the Web of Science (WoS) database to collect data on journal articles related to "big data" published between 2002 and 2016. Key techniques included:
- Bibliometric Analysis: Standard methods to analyze publication trends, journal distributions, author contributions, and institutional collaborations.
- Co-word Analysis: To identify research topics and frontiers by analyzing the co-occurrence of keywords.
- Document Co-citation Analysis (DCA) & Author Co-citation Analysis (ACA): To understand the intellectual structure and identify influential works and researchers by analyzing how often documents or authors are cited together.
- Information Visualization: Using tools like UCINET and CiteSpace V to create visual representations of the complex relationships within the data, making it easier to understand networks, structures, and trends.
Datasets
The study utilized data from the Web of Science (WoS) database, specifically focusing on journal articles. The dataset comprised:
- 4927 effective journal articles
- Published between 2002 and 2016
- From 1729 journals
- Contributed by 16,404 authors
- From 4137 institutions
No new datasets were introduced; the focus was on analyzing existing literature.
Key Insights and Ideas
- Exponential Growth: Big data research experienced a dramatic surge, particularly from 2010 onwards, indicating a rapid maturation and increasing global interest.
- Collaborative Ecosystem: While individual contributions are important, the field is increasingly collaborative, with an average of 3-4 authors per paper in recent years.
- Key Players Emerge: Specific journals (like PLoS One, Nature, Science) and institutions (Chinese Academy of Sciences, Tsinghua University, US universities) are consistently at the forefront, acting as hubs for knowledge dissemination and creation.
- Australia's Early Strength: Australia showed a notable early strength in big data research, particularly in terms of prolific authors.
- Emerging Hotspots: Beyond core topics like data mining and machine learning, the research is expanding into areas like data science (privacy, management), deep learning, and applications in social networks, smart cities, and healthcare.
- "Matthew Effect" in Play: Similar to other scientific fields, big data research exhibits a "Matthew effect," where a small percentage of journals, institutions, and authors contribute a disproportionately large share of publications.
Key Results
- Publication Trends: A slow start from 2002-2009, followed by an exponential growth spurt from 2010-2016, with annual articles increasing 600-fold.
- Author Growth: The number of authors also saw a similar exponential rise, reaching over 9,500 in 2016.
- Core Journals: PLoS One led in publications, while Nature and Science were highly co-cited, indicating their foundational role.
- Top Institutions: Chinese Academy of Sciences and Tsinghua University were leading in publications, though their betweenness centrality suggested a less dominant "pivotal" role compared to some US institutions.
- Core Authors: Rajiv Ranjan was identified as a central node in the author collaboration network. Jeffrey Dean emerged as the most co-cited author.
- Keywords: "Big Data" was the most frequent keyword, followed by "cloud computing," "machine learning," and "MapReduce." Emerging topics included "data science," "deep learning," and "optimization."
Limitations and Future Work
- Data Source Limitations: The study relied solely on the Web of Science, which might not capture all relevant publications, especially from certain regions or conference proceedings.
- Author Name Ambiguity: The authors acknowledge the challenge of disambiguating author names and affiliations, which can affect the accuracy of author-centric analyses.
- "Matthew Effect" Interpretation: While the "Matthew effect" is observed, the authors note that the research landscape is still somewhat fragmented, with a large number of authors not considered "core."
- Future Work: The paper suggests that as big data research continues to evolve, collaboration networks will likely strengthen. Future research could delve deeper into specific subfields, analyze the impact of emerging technologies, and explore the ethical and societal implications of big data for sustainable development. The identified challenges (data heterogeneity, processing efficiency, security, energy consumption) also point towards crucial areas for future research and development.
Summary updated