报告时间：2026年11月10日下午13:30

报告地点：北湖东校区数统新楼201室

主办单位：伟德体育

报告人：孙玉莹

报告人简介：孙玉莹，中国科伟德体育数学与系统科学研究院研究员、博士生导师，主持国家优秀青年科学基金项目，入选北京市青年重点项目（原北京市杰青），入选中国科协“青年人才托举工程”。研究兴趣主要有计量经济学、经济预测理论与方法等。在Management Science, Journal of Econometrics, Journal of Business & Economic Statistics，《中国科学：数学》，《管理科学学报》等国内外重要期刊上发表论文30余篇；基于新理论与新方法，作为主笔，30余篇政策报告获国家领导人批示或被中办、国办采用。先后获成思危优秀科研成果奖、陈景润未来之星、关肇直青年研究奖、中国科伟德体育数学与系统科学研究院“重要科研进展奖（2017，2019）”等。

摘要：In the evolving landscape of digital healthcare, efficient online medical triage systems are central to platform operations and service quality. This study proposes the Penalized Ensemble for Multinomial Logistic Learning (PEMLL), an ensemble learning framework for high-dimensional triage prediction that combines penalization with model averaging to address feature complexity and model uncertainty in multinomial settings. We establish the asymptotic optimality and consistency of PEMLL under general conditions and evaluate its performance through simulation studies and a real-world application using patient data from JD Health, a leading online healthcare platform in China. Empirically, PEMLL improves predictive accuracy upon standard model selection and machine learning methods. Using a model-based simulation that maps misclassification into additional waiting time and projected revenue loss, PEMLL is associated with lower operational burden and economic loss than competing approaches. These findings illustrate how improvements in prediction accuracy correspond to meaningful operational and economic gains in platform-based service systems, highlighting the role of AI-driven predictive models in supporting decision-making.