基于大数据的舒缓类化妆品安全性分析：消费者不良反应挖掘与成分风险分析

Big Data-Driven Safety Analysis of Soothing Cosmetics: Mining of Consumer-Reported Adverse Reactions and Ingredient Risk Profiling

摘要: 目的：舒缓类化妆品市场需求量持续增长，本研究旨在利用互联网公开数据系统分析国内网售舒缓类化妆品的市场特征与安全性现况。方法：收集2016年至2021年国内电商平台上宣称具有舒缓功效的化妆品数据，涵盖销量、剂型、产地、价格、成分及用户评论等。利用自然语言处理技术进行虚假评论过滤、情感分析和关键词提取，识别疑似不良反应信号，并探讨不良反应率与各因素之间的相关性。结果：共筛选出181个产品。润肤水、精华和防晒霜为消费者购买最多的剂型。国产品牌占比44.50%，居于首位。不良反应率与产品成分总数及功效成分数量无显著相关性，与风险成分数量及风险成分占比呈正相关，与功效成分占比呈负相关。识别出13种需重点关注的风险成分，部分低风险成分在高不良反应产品中频繁出现，提示其安全性在中国人群中需重新评估。结论：舒缓类化妆品的安全性不取决于成分数量或价格，而受风险成分占比和功效成分浓度的综合影响。消费者应基于多维信息进行选购判断。

Abstract: Objective: The market demand for soothing cosmetics continues to grow. This study aimed to systematically analyze the market characteristics and safety profile of soothing cosmetics sold online in China using publicly available internet data. Methods: Data on cosmetics with soothing claims were collected from Chinese e-commerce platforms between 2016 and 2021, encompassing sales volume, formulation type, country of origin, price, ingredient lists, and consumer reviews. Natural language processing techniques were applied for fake review filtering, sentiment analysis, and keyword extraction to identify suspected adverse reaction signals. Associations between the suspected adverse reaction rate and various product attributes were examined. Results: A total of 181 products were included. Lotions, essences, and sunscreens were the most frequently purchased formulation types. Domestic Chinese brands accounted for 44.50% of the sample, ranking first among all countries of origin. The suspected adverse reaction rate showed no significant correlation with the total number of ingredients or the number of functional ingredients, but was positively correlated with both the number and proportion of risk ingredients, and negatively correlated with the proportion of functional ingredients. Thirteen risk ingredients warranting prioritized attention were identified. Certain low-risk ingredients appeared disproportionately in products with high adverse reaction rates, suggesting that their safety in the Chinese population merits reassessment. Conclusion: The safety of soothing cosmetics is determined primarily by the proportion of risk ingredients and the concentration of functional ingredients, rather than by the number of ingredients or product price. Consumers should base their purchasing decisions on multidimensional information.