讲座题目: Research on Dynamic Monitoring and Diagnosis Methods for Automobile Quality Based on Online Complaint Text Analysis
讲座摘要: The automobile industry is one of the important pillar industries of the national economy, and the issue of automobile quality and safety has attracted much attention. With the development of the Internet, more and more automobile users choose to post comments on automobile quality issues online. Reasonable use of online complaint texts about quality problems can help automobile enterprises achieve effective automobile quality monitoring. The existing researches mainly focus on the static extraction of automobile quality defects, such as defect identification and classification, and seldom study the dynamic change of automobile quality problems. Therefore, according to the characteristics of automobile quality complaint texts on the Internet, this paper firstly applies the Latent Dirichlet Allocation model to extract the quality defect topics, and classifies the complaint texts by topics. Then calculate the time intervals and sentiment scores of complaint texts under each defect topic, so that the change trend of complaints can be described from two dimensions of frequency and intensity simultaneously. Secondly, based on rank test and empirical Copula, a two-dimensional robust control chart is proposed to monitor the time interval, sentiment score and their correlation online at the same time, and further identify the abnormal causes after the control chart alarms, so as to realize real-time monitoring and diagnosis of dynamic changes in automobile quality. Finally, we take Volkswagen Sagitar as an example, select its online complaint texts from 2010 to 2021 on the Automobile Quality Website to illustrate the application of the proposed method in practice.
讲座时间:2023年9月21日(周四)上午9:30
讲座方式:线上举办(腾讯会议322-716-113)
嘉宾简介:宋贽,博士,副教授,中国现场统计研究会可靠性工程分会理事,现任沈阳农业大学理学院副院长。主要从事统计质量控制、服务过程统计建模、社交媒体数据分析等领域的研究。主持国家自然科学基金青年基金1项,教育部人文社会科学研究青年基金1项,教育部产学合作协同育人项目2项,省级科研课题4项,省级教改课题 1项。辽宁省线下一流课程“概率论”负责人。发表SCI检索论文10余篇。授权发明专利、软件著作权2项。