有关抑郁症的机器学习模型

Machine Learning Models for Depression

摘要: 为了解近5年来有关抑郁症的相关机器学习模型。检索2020年至2025年Web of Science数据库核心合集中有关抑郁症的机器学习模型的相关文献。采用CiteSpace 6.3R1对其发文量、国家、机构、作者、共现关键词、关键词聚类和时间线谱进行可视化分析。共收录文章1282篇，近5年来年度发文量呈现总体上升趋势。对国家、机构和作者的分析表明，中国是发文量最多的国家；目前尚未形成核心作者群，Hu Bin为代表的团队共发表了22篇文章，是最核心的团队；中国科学院是该网络最核心的机构。对关键词的共现与聚类图谱进行分析，当前研究热点主要围绕“研究靶点与特定人群”、“多模态数据源”以及“核心算法与模型评估”三大维度展开。对最近爆发词和时间线谱分析表明，抑郁症机器学习模型领域已从2020年左右的技术爆发期，过渡到当前的方法论深化、多模态融合框架构建和临床精细化应用验证的探索阶段。

Abstract: This paper aims to investigate the development and research trends of machine learning models related to depression over the past five years. Relevant literature published from 2020 to 2025 was retrieved from the Web of Science Core Collection. CiteSpace 6.3R1 was used to conduct a visual analysis of publication volume, countries, institutions, authors, keyword co-occurrence, keyword clustering, and timeline views. A total of 1282 papers were included, with the annual publication volume showing an overall upward trend in the past 5 years. Analysis of countries, institutions, and authors showed that China was identified as the country with the highest publication volume. Although a stable core author group has not yet fully formed, the team represented by Hu Bin was the most prominent (22 papers), and the Chinese Academy of Sciences was the central institution in the network. Analysis of keyword co-occurrence and clustering indicates that research hotspots are currently centered around three key dimensions: “research targets and specific populations”, “multimodal data sources”, and “core algorithms and model evaluation”. The analysis of burst terms and timeline views suggests that the field has transitioned from a phase of rapid technical emergence around 2020 to a current exploratory stage focused on methodological deepening, the construction of multimodal fusion frameworks, and refined clinical application validation.