
Machine Learning in Predicting Extreme Earnings and Abnormal Returns
主讲嘉宾 康健
报告人简介:
康健博士现任东北财经大学金融学院讲师(准聘副教授)。他的研究兴趣集中于计量经济学、金融与会计的交叉领域,涵盖气候变化与风险、资本市场及公司披露等方向。他在Journal of Econometrics、Energy Economics和Econometrics等国际知名期刊上发表了多篇学术论文。此外,他担任Journal of Accounting, Auditing, and Finance、China Accounting and Finance Review等国际期刊,以及美国会计协会年会等国际会议的匿名审稿人。
报告论文摘要:
This paper uses machine learning techniques to predict a specific type of earnings changes (i.e., one-year-ahead earnings jumps and crashes). We focus on extreme changes in earnings as market participants are more interested in such changes. More importantly, machine learning methods are better able to predict a binary variable than a continuous variable. Overall, we find that a random forest model outperforms linear models in predicting earnings jumps and crashes. Moreover, we show that over a 12- month holding period, a hedge portfolio that uses the top 5% predicted earnings jump and crash probabilities derived from a random forest model generates significant annual size-adjusted returns of 14.99%. We also find that long-short portfolios based on predicted earnings jump or crash probability generate abnormal returns around earnings announcements. In general, our findings suggest that a random forest model is useful in predicting extreme earnings events and predicted probabilities help investors make profitable investment decisions.
活动时间: 2025年4月3日 星期四 上午9:30
活动地点: 金融学院会议室(劝学楼345小会议室)
主办单位:金融学院、实验经济学实验室欢迎广大教师、同学参加!