机器学习在微生物生态领域的应用文献计量分析

Bibliometric analysis of machine learning in microbial ecology

摘要:

【背景】 随着新型台站和测序方法的迅猛发展，微生物生态学数据已经呈现出爆炸式的增长，为理解微生物在全球生态系统中的功能提供了坚实的基础。然而，这些庞大的数据集的处理和分析对传统方法来说是一个挑战，机器学习因其在处理大数据方面的优势而成为解决这一问题的关键技术。【目的】 基于文献计量学，本文全面探索机器学习在微生物生态学研究中的应用，包括其发展趋势、现状及热点，为未来机器学习和微生物生态研究的结合指明方向。【方法】 获取并分析Web of Science (WOS)核心合集数据库中1991–2023年间发表的相关文献，运用可视化软件CiteSpace探究发文量演变特征、国际合作情况及学科交叉现状，利用Carrot2对文本数据进行挖掘，构建可视化知识图谱。【结果】 机器学习在微生物生态学研究中应用的数量以2018年为界经历了稳定增长和暴发增长两个时期，相关应用正逐渐成为各国研究的热点，研究成果持续增长，相关学科之间的交叉融合越来越紧密，特别是微生物生态学与化学、物理、环境、计算机等学科之间的合作，为科学研究的进展提供了新的视角。机器学习在微生物生态学中的应用广泛。在早期，研究主要集中在序列识别和物种分类上。自2018年以来，随着深度学习和计算机视觉技术的发展，研究焦点转向复杂系统的预测。这两个时期关键词的对比展示了机器学习技术在微生物生态学中的应用从基础的数据处理和分析逐渐转向更加复杂、高级的预测模型。【结论】 基于文献计量分析结果，结合机器学习在微生物生态中应用的数据缺乏、模型选择困难和可解释性差等挑战，后续研究应更加重视国际合作和数据共享，加强学科交流，推动可解释机器学习的发展。

Abstract:

[Background] With the rapid development of new types of stations and sequencing methods, the data in microbial ecology has experienced explosive growth, providing rich resources for revealing the role of microorganisms in global ecosystems. However, the processing and analysis of these large datasets pose challenges to conventional methods. Machine learning, with unique advantages in handling big data, has become a key technology to address these challenges. [Objective] This study comprehensively explored the developmental trends, current status, and hotspots in the application of machine learning in microbial ecology through bibliometric analysis, aiming to guide the future integration of machine learning with the research of microbial ecology. [Methods] The relevant articles published between 1991 and 2023 in the Web of Science (WOS) Core Collection were collected and analyzed. CiteSpace was used to visualize the evolution of the number of publications, international collaboration, and interdisciplinary status. Carrot2 was employed to mine textual data and build knowledge maps. [Results] The application of machine learning in the research of microbial ecology has undergone two distinct phases: stable growth followed by explosive growth, with 2018 marking a turning point. This research field has gained increasing attention, which led to continuous growth in the research output. The integration of machine learning with other disciplines, especially chemistry, physics, environmental sciences, and computer science, has become increasingly tight-knit, providing new perspectives for the advancement of scientific research. The application of machine learning in microbial ecology is extensive. The early studies primarily focused on sequence recognition and species identification. However, as deep learning and computer vision technologies have kept advancing since 2018, the research focus shifted towards predicting complex systems. The comparison of keywords between the two phases highlighted the evolution of machine learning in microbial ecology from basic data processing and analysis to complex and advanced prediction models. [Conclusion] According to the results of bibliometric analysis and considering challenges like data scarcity, difficulties in model selection, and poor interpretability in applying machine learning in microbial ecology, we suggest that international collaboration, data sharing, and interdisciplinary exchange should be emphasized in the future to promote the development of interpretable machine learning.