Monday, April 12, 2021 10am to 11:30am
About this Event
Yichen Cheng (Insight, Assistant Professor) https://robinson.gsu.edu/profile/yichen-cheng/
Houping Xiao (Insight, Assistant Professor) https://robinson.gsu.edu/profile/houping-xiao/
We will present and discuss two novel applications of machine learning in the capital markets.
In the first project (by Sean Cao, Xuxi Guo (Ph.D. student), Houping Xiao, and Baozhong Yang), we consider how ML can be used to understand human decisions. We employ deep learning models to study analyst forecasts. Our model is capable of learning nonlinear interactions among analyst and forecast features and identify ex-ante skilled ML star analysts. The ML-identified star analysts predict earnings more accurately than human-labeled star analysts. As an application, we aggregate the predictions by skilled analysts to form a ML-based forecast consensus that can provide significant abnormal returns to investors. The methods we develop can also be applied to other settings of human forecasts and information aggregation such as online forums, political opinions, and macroeconomic outlooks.
In the second project (by Sean Cao, Yichen Cheng, Yusen Xia, and Baozhong Yang), we apply ML to analyze unstructured, high-dimensional image data from corporate executive presentations. Our model applies state-of-the-art deep learning models to extract forward-looking information from firms’ business operation graphics. We validate the forward-looking visual information using market reactions and analyst revisions. While institutional investors and skilled analysts also react to such information, our model would help to level the playing field by providing valuable information to all investors. Our approach also provides a deep learning toolbox specifically designed for extracting value-relevant information from business images.
Register in advance for this meeting: https://rcbdeanstemp.wufoo.com/forms/z1x7xngi08zbl1i/
Those who register will receive, in advance of the meeting, a link and information about joining the event.
///////////////////////////
RCB Research Coffee Breaks:
These are opportunities to spend time with research colleagues from across the college (and university) in a format that provides for exchange of ideas, getting to know people from other departments, and generally increasing collegial interaction that reaches across the silos. The focus and format of each meeting may vary, but generally, each will focus on active research projects that could benefit from collegial input, thematic discussions of college/university opportunities or challenges, or other topics. The sessions are offered in two formats: 1) A one-hour discussion format—this includes open conversation focused on the project(s) or topic(s) of the day. The expectations of lead participants are small: The sessions are low key, informal, and in keeping with the relatively informal and unstructured format, there will be a rarely-waived NO PowerPoint rule (please check your PowerPoints at the door upon entry). If one must use a visual or handout—there is a limit to one page (so you can hold your coffee in one hand and the page in the other!). 2) For projects/topics that are better suited for a full presentation, a 1.5-hour traditional seminar format is offered. These will maintain the constructive, somewhat informal culture of our coffee discussions, but will allow for a full presentation focused on research ideas or projects that could inform and engage colleagues while benefitting from their collegial input (and yes, PowerPoints are welcome!). Both college-level event types are intended for topics that are cross-cutting and that will be relevant to multiple departments. Both can help to bridge otherwise disconnected areas, engage expertise from different perspectives, and help to make a big college/university populated by relative strangers seem smaller -- and occupied by friends and colleagues.
0 people are interested in this event