Abstract: This study focuses on relevant content in the field of automotive intelligent driving on WeChat public accounts over the past three years, using "automotive intelligent driving" and "automotive industry" as keywords for thematic retrieval. After data cleaning and organisation, 121 high-quality texts were selected as the analysis corpus. The study employs the latent dirichlet allocation (LDA) topic model to train the obtained texts, identifying the core research hotspots in China's intelligent driving technology field, namely computing power, risk, risk level, data models,driver monitoring, assisted driving, sensor fusion, and intelligent vehicles. On this basis, the study further points out that automotive industry practitioners and learners in related fields should pay sufficient attention to eight aspects–computing power research, risk research, risk level research, data model research, driver monitoring research, assisted driving research, sensor fusion research, and intelligent vehicle research–when conducting research in intelligent driving technology in the future. For hotspot theme task identification, the study verifies the effectiveness and practicality of the LDA topic model through K-means clustering. Its analysis results are conducive to researchers and industry personnel quickly grasping frontier developments and research focuses in the field. Additionally, the study's conclusions can provide references for automotive industry policy-making and corporate technology planning, helping the automotive sector form a clearer development path in key intelligent driving technologies.

Key words: LDA topic model; automotive intelligent driving technology; hot topics; K-means clustering; development path

摘要： 文章以微信公众号中近 3 年汽车智驾领域的相关内容为研究对象，以“汽车智驾”和 “汽车产业”为关键词进行主题检索，经过数据清洗和整理，最终选取 121 条高质量文本内 容作为分析语料。研究通过隐含狄利克雷分配（LDA）主题模型对所获取的文本进行训练， 得到国内汽车智能驾驶技术领域的核心研究热点，包括算力、风险与风险程度、数据模型、 驾驶员监测、辅助驾驶、传感器融合及智能汽车。在此基础上，研究进一步指出，汽车行业 从业者及相关领域学习者在未来开展智能驾驶技术研究时，应在算力研究、风险研究、风险 程度研究、数据模型研究、驾驶员监测研究、辅助驾驶研究、传感器融合研究、智能汽车研 究八个方面给予足够关注。在热点主题任务识别上，研究通过 K-means 聚类验证了 LDA 主题 模型的有效性和实用性，其分析结果有利于研究者及本行业人员快速掌握领域内的前沿动态 与研究焦点。同时该研究结论也可为汽车产业政策制定、企业技术布局提供参考，助力汽车 行业在智能驾驶关键技术领域形成更清晰的发展路径。

关键词: LDA 主题模型；汽车智能驾驶技术；热点主题；K-means 聚类；发展路径