联系我们

论文查询结果	返回搜索

论文编号：	15371
作者编号：	2120223804
上传时间：	2025/6/8 0:06:57
中文题目：	基于BERTopic和HCAT的医疗投诉信息主题建模与分类优化
英文题目：	Research on Topic Modeling and Classification of Healthcare Complaint Information Based on BERTopic and HCAT
指导老师：	徐曼
中文关键字：	医疗投诉；BERTopic；HCAT；医疗质量管理；主题建模
英文关键字：	Healthcare Complaints;BERTopic;HCAT;Healthcare quality management;Topic modeling
中文摘要：	随着医疗体制改革的深入，患者体验管理已成为医疗服务质量控制的核心维度。医疗投诉数据作为医患互动的真实映射，蕴含着优化医疗服务的重要线索。然而，传统的人工分类方法存在效率低下、主观性强、难以识别新兴问题等局限性，难以满足精细化管理需求。为此，本文构建了数据驱动的医疗投诉智能分析框架，突破人工分类的效能瓶颈，提升投诉分析的精准性与时效性。 本研究基于天津某三甲医院 2022-2023 年全渠道投诉数据，结合改进BERTopic 主题建模技术与医疗投诉分析工具（HCAT），构建“语义解析—动态分类—决策支持”闭环体系。首先，采用文本分析法进行自动化处理，构建医疗专用词典，优化数据预处理流程，以提升医疗文本的分析准确性。其次，针对主题建模方法的局限性，创新性融合 Sentence-BERT 深度语义嵌入、PCA 降维与 GMM 软聚类算法，优化BERTopic模型，以提高主题提取的精准性，并降低传统 HDBSCAN 算法的离群值敏感性。为增强分类标准化，本研究构建主题-HCAT 映射规则库，实现25个投诉主题簇向6类管理维度的自动化归类。最后，结合比较研究法，横向评估不同主题建模方法和投诉分析框架，并通过纵向时间序列分析揭示投诉主题的演化趋势，结合政策变动预测未来可能的新兴投诉热点，为医疗管理者提供趋势预警和决策支持。 研究结果表明，改进后的 BERTopic 模型与 HCAT 框架结合后，有效提升了医疗投诉主题识别精度，增强了自动化分类的稳定性与可解释性，并通过智能化分析手段，为医疗机构提供科学、前瞻的决策支持。研究不仅提升了 BERTopic在中文医疗投诉分析中的适配性，同时验证了 HCAT 在我国医疗机构的应用潜力，为医疗投诉智能化分析提供了新思路。研究成果可为医疗质量改进、患者投诉管理优化及风险预警提供数据支撑，并为医院精细化管理提供决策参考。
英文摘要：	With the ongoing healthcare system reforms, patient experience management has become a core component of healthcare quality control. Medical complaint data, reflecting real interactions between patients and healthcare providers, holds key insights for improving healthcare services. However, traditional manual classification methods suffer from inefficiencies, subjectivity, and a failure to identify emerging issues, which makes them inadequate for supporting the fine-grained management required. To ad- dress this, this study develops a data-driven intelligent analysis framework for medical complaints, overcoming the limitations of manual classification and enhancing the precision and timeliness of complaint analysis. This study is based on a comprehensive dataset of medical complaints from a top- tier hospital in Tianjin, spanning 2022 to 2023. It integrates an improved version of the BERTopic topic modeling technique with the Healthcare Complaints Analysis Tool (HCAT) to create a closed-loop system of “semantic parsing—dynamic classification —decision support. ”First, we employ text analysis methods to automate data processing, creating a specialized medical dictionary to optimize the data preprocessing flow and improve the accuracy of medical text analysis. Second, addressing the limitations of traditional topic modeling techniques, we innovate by integrating Sentence-BERT for deep semantic embedding, PCA for dimensionality reduction, and GMM soft clustering algorithms to enhance the BERTopic model, improving topic extraction accuracy and reducing the sensitivity to outliers seen in traditional HDBSCAN algorithms. To standardize classification, we develop a topic-to-HCAT mapping rule library, which enables the automated categorization of 25 complaint topic clusters into 6 management dimensions. Finally, using comparative research methods, we evaluate different topic modeling techniques and complaint analysis frameworks, and through longitudinal time-series analysis, we reveal the evolution of complaint topics, predicting emerging complaint hotspots in light of policy changes and providing trend alerts and decision support for healthcare managers. The results show that the combination of the improved BERTopic model and the HCAT framework significantly enhances the precision of medical complaint topic identification, strengthens the stability and interpretability of automated classification, and provides healthcare institutions with scientific, forward-looking decision support through intelligent analysis. This research not only improves the applicability of BERTopic for analyzing Chinese medical complaints but also validates the potential of HCAT for use in Chinese healthcare institutions, offering new insights into the intelligent analysis of medical complaints. The findings provide valuable data support for healthcare quality improvement, patient complaint management, and risk forecasting, offering important decision-making references for hospitals aiming to refine their management practices.
查看全文：	预览  下载(下载需要进行登录)