消费品投诉识别与缺陷特征发现 —— 基于 BERTopic 模型

摘要：

【目的】为改进消费品质量和服务、提升消费者满意度和品牌竞争力提供依据，助力消费品市场有序健康发展。【方法】以2016—2021年从政府官方平台获取的23,736条投诉文本为基础，应用BERTopic模型进行主题挖掘，识别主要主题及演化趋势，并借助投诉主题网络分析缺陷特征之间的关联性。【结果】消费品投诉主要集中在电子产品的质量缺陷和售后服务不足，消费者对手机电池问题导致的自动关机关注度较高，投诉主题网络揭示了投诉内容之间的复杂关系，维修、售后等问题在网络中占据桥梁地位，消费者重视品牌，并对品牌和产品类别的投诉表现出显著差异。【结论】所构建的主题模型能够较好地识别投诉的类别、品牌、缺陷特征等，并详细展示了各投诉内容之间的关联性。后续将进一步扩大研究样本范围，从更广泛的渠道获得投诉文本并进行信息挖掘。

Abstract:

[Objective] This study aims to provide a basis for improving the quality and service of consumer goods, enhancing consumer satisfaction and brand competitiveness, and promoting the orderly and healthy development of the consumer goods market. [Methods] Utilizing 23,736 consumer complaint texts obtained from government official platforms between 2016 and 2021, this study applies the BERTopic model for topic mining to identify key themes and their evolutionary trends. Furthermore, a network analysis of complaint topics is conducted to examine the associations between defect characteristics. [Results] The analysis reveals that consumer complaints predominantly focus on quality defects in electronic products and inadequate after-sales service. Consumers exhibit a high level of concern regarding automatic shutdown issues caused by mobile phone battery problems. The complaint topic network uncovers complex relationships within complaint content, highlighting that issues related to repair and after-sales service serve as bridging nodes in the network. Additionally, consumers demonstrate a strong emphasis on brands, with notable differences in complaint patterns across brands and product categories. [Conclusion] The constructed topic model effectively identifies complaint categories, brands, and defect characteristics while elucidating the intricate relationships among different complaint contents. Future research will expand the sample scope and incorporate complaint texts from broader sources to enhance information mining. 

引用格式：田晶晶，洪瑶瑶，韦海霞，等.消费品投诉识别与缺陷特征发现—— 基于BERTopic模型[J].标准科学，2025(6)：116-125.

基金项目：本文受国家发展和改革委员会社会治理和智慧社会科技支撑重点项目“消费者投诉举报信息智能挖掘关键技术研究及应用示范”（项目编号：2024YFC3307800）；国家市场监督管理总局科技计划项目“缺陷汽车产品召回过程追溯关键技术研究及应用”（项目编号：2024MK188）资助。 

作者简介：田晶晶，博士，副研究员，研究方向为缺陷产品召回信息管理、数据挖掘应用等。 洪瑶瑶，通信作者，博士研究生，研究方向为缺陷产品召回、博弈论、大语言模型、机器学习等。 韦海霞，硕士，高级工程师，研究方向为缺陷汽车召回。 刘成伟，硕士，高级工程师，研究方向为缺陷汽车质量管理。 林春伟，硕士，高级工程师，研究方向为缺陷汽车质量控制。