报告题目:Homogenization Dilemma of Quantitative Trading
报告人: 刘彦初 教授(中山大学)
时间:2026年10月22日 15:30
地点:管理学院A北204
报告内容摘要:
We study algorithmic homogenization arising from the widespread use of machine-learning models in asset pricing. We construct a simulated ecosystem of independent quantitative strategies and track the evolution of strategy behavior from 2018 to 2024 using approxi-mately 266 million out-of-sample stock-level return forecasts generated from 6.1 million stock-day observations. The experiments show that extensive use of complex models by as-set managers does not intensify financial risk; concentrated use of interpretable models is in-stead an important source of risk accumulation. Linear models are more likely to form high-ly overlapping holdings, thereby concentrating trading pressure, whereas neural networks exhibit lower overall holding similarity and, to some extent, mitigate synchronized trading pressure across strategies. A healthy and resilient quantitative-trading ecosystem therefore depends on maintaining diversity in model architectures, thereby mitigating the homogeni-zation risk created by the concentrated use of interpretable models.
【刘彦初个人简介】
刘彦初博士现就职于中山大学岭南学院,担任党委委员、副院长、金融学教授、博士生导师,并担任中山大学教育发展与校友事务办公室主任、中山大学资本市场研究院院长、中山大学粤港澳发展研究院教授(双聘)、国际学术期刊《Digital Finance》副主编(Associate Editor)。主要研究兴趣为数智金融。在上述研究领域取得多项成果,发表于《管理科学学报》(*4)、《Operations Research》(*2)、《INFORMS Journal on Computing》、《Journal of Banking and Finance》、《Journal of Futures Markets》(*7)、《Journal of Economic Dynamics and Control》(*2)、《European Journal of Operational Research》(*2) 等国内外主流学术期刊。作为负责人主持国家自然科学基金(*3),中国期货业协会研究课题(*2),上海期货交易所年度课题,中国金融期货交易所年度课题,郑州商品交易所期货及衍生品研究所对外合作课题,广州期货交易所对外合作课题(*2),证监会广东监管局专项课题,广东省委有关部门委托课题,国家高端智库重点课题,头部期货公司委托课题等多项科研基金。多篇咨政建议获得中央和广东省委有关部门采纳。
欢迎有兴趣的教师,全体博士生、硕士生参加。
管理学院科研办公室
2026年10月19日