基于情感分类和主题建模的患者体验分析

Patient Experience Analysis based on Sentiment Classification and Topic Modeling

目的分析患者体验文本,改善医疗服务质量,提升患者满意度,为大量患者体验文本提供有效的分析手段。方法对患者体验文本使用长短期记忆模型进行情感分类;对负向体验文本利用主题生成模型与点互信息进行主题分析。结果情感分类达到平均94.66%的精确率、94.77%的召回率与94.61%的F1值,提炼出负向患者体验文本包含的主题,并分析了各年度主题热度与变化趋势。结论该方法可以有效的识别负向患者体验文本,并进一步对负向体验文本的主题进行提炼与分析,有助于医院管理者从大量患者体验文本中及时了解患者感受,挖掘服务问题,提升管理效率。

Objective Analyze patient experience, improve the quality of medical services and patient satisfaction, and provide effective analysis tools for huge volume of patient experience texts. Methods Long-term and Short-term Memory models were used to classify the emotion of patient experience texts. The negative experience texts was analyzed by topic generation model and point mutual information. Results The sentiment classification achieved an average accuracy of 94.66%, a recall rate of 94.77%, and an F1 value of 94.61%. Topics in the negative patient experience texts were exacted, annual topic heats and trends were analyzed. Conclusion This method can effectively identify negative patient experience, and further refine and analyze the topics of negative experience, which is helpful for hospital administrators to understand patient feelings, explore service problems, and improve management efficiency.