Liu Jingfang, Lu Shuangjinhua, Lu Caiying
School of Management, Shanghai University, Shanghai 200444, China.
Healthcare (Basel). 2021 Oct 12;9(10):1353. doi: 10.3390/healthcare9101353.
(1) Background: The COVID-19 pandemic is globally rampant, and it is the common goal of all countries to eliminate hesitation in taking the COVID-19 vaccine and achieve herd immunity as soon as possible. However, people are generally more hesitant about the COVID-19 vaccine than about other conventional vaccines, and exploring the specific reasons for hesitation with the COVID-19 vaccine is crucial. (2) Methods: this paper selected text data from a social platform to conduct qualitative analysis of the text to structure COVID-19 vaccine hesitancy reasons, and then conducted semiautomatic quantitative content analysis of the text through a supervised machine-learning method to classify them. (3) Results: on the basis of a large number of studies and news reports on vaccine hesitancy, we structured 12 types of the COVID-19 vaccine hesitancy reasons. Then, in the experiment, we conducted comparative analysis of three classifiers: support vector machine (SVM), logistic regression (LR), and naive Bayes classifier (NBC). Results show that the SVM classification model with TF-IDF and SMOTE had the best performance. (4) Conclusions: our study structured 12 types of COVID-19 vaccine hesitancy reasons through qualitative analysis, filling in the gaps of previous studies. At the same time, this work provides public health institutions with a monitoring tool to support efforts to mitigate and eliminate COVID-19 vaccine hesitancy.
（1）背景：新冠疫情在全球肆虐，消除对新冠疫苗接种的犹豫并尽快实现群体免疫是各国的共同目标。然而，人们对新冠疫苗的犹豫普遍高于其他传统疫苗，探究对新冠疫苗犹豫的具体原因至关重要。（2）方法：本文从一个社交平台选取文本数据进行文本定性分析，以构建新冠疫苗犹豫原因，并通过监督机器学习方法对文本进行半自动定量内容分析以对其进行分类。（3）结果：在大量关于疫苗犹豫的研究和新闻报道基础上，我们构建了12种新冠疫苗犹豫原因。然后，在实验中，我们对支持向量机（SVM）、逻辑回归（LR）和朴素贝叶斯分类器（NBC）这三种分类器进行了对比分析。结果表明，采用词频逆文档频率（TF-IDF）和合成少数过采样技术（SMOTE）的支持向量机分类模型性能最佳。（4）结论：我们的研究通过定性分析构建了12种新冠疫苗犹豫原因，填补了以往研究的空白。同时，这项工作为公共卫生机构提供了一种监测工具，以支持减轻和消除新冠疫苗犹豫的努力。
