欢迎光临武汉大学经济与管理学院!
English Version设为首页加入收藏联系方式
会议专题 首页 - 会议专题 - 正文
景林珞珈金融论坛(第196-202期)——2021春季金融科技Workshop
时间:2021-04-15    点击数:

Jinglin Luojia Finance Seminar: Spring FinTech Workshop, 2021

 

日期:2021年4月24日(星期六)

场地:经管院A421

主办单位:

    武汉大学经济与管理学院金融系

    武汉大学金融科技研究中心

    武汉大学大数据研究院金融大数据研究中心

 

 

9:00 – 9:50

1. Different Strokes: Return Predictability Across Stocks and Bonds with Machine Learning and Big Data

基于大数据与机器学习的股票与债券市场预测

报告人:姜富伟,中央财经大学

9:50 – 10:40

2.基于自编码机器学习的金融大数据资产定价研究

报告人:唐国豪,湖南大学

10:40 – 10:50

茶歇

10:50 – 11:40

3. Delta Hedging and Volatility-Price Elasticity: A Two-Step Approach

报告人:杨学伟,南京大学

11:40 – 13:30

午餐:珞珈山庄

14:00 – 14:50

4.Labor Hiring and Stock Returns

报告人:陈坚,厦门大学

14:50 – 15:40

5. Media-expressed Tone, Option Characteristics, and Stock Return Predictability

报告人:刘彦初,中山大学

15:40 – 15:50

茶歇

15:50 – 16:40

6. Completing the Market: Generating Shadow CDS Spreads by Machine Learning

报告人:李剑,浙江工商大学

16:40 – 17:30

7. Selecting Mutual Funds from the stocks they hold: A machine learning approach

报告人:李斌,武汉大学

晚餐:TBD

 

报告人及文章摘要

1.姜富伟

个人简介:姜富伟,教育部青年长江学者,中央财经大学金融工程系主任,教授、博士生导师。研究领域包括资产定价、行为金融、金融科技、金融大数据与机器学习等,在金融学顶级期刊Journal of Financial Economics、Review of Financial Studies、Management Science及《金融研究》、《经济学季刊》、《管理科学学报》等发表论文30余篇。主持国家自然科学基金等课题项目5项,参与课题8项。被评为ESI经济管理类全球前1%最高被引用论文、RFS最高被引用论文、《世界经济年鉴》最佳国际金融学和最佳世界经济统计学论文。被《哈佛商业评论》、《哈佛法学院公司治理与金融监管论坛》、《清华金融评论》、《CFA文摘》、招商证券、瑞士银行等转载应用。获《金融研究》优秀论文奖、国际金融管理协会年会最佳论文奖、亚洲金融协会年会最佳论文奖等学术奖励。

题目:基于大数据与机器学习的股票与债券市场预测(Different Strokes: Return Predictability Across Stocks and Bonds with Machine Learning and Big Data)

摘要:We investigate the return predictability across stocks and bonds using big data and machine learning. We find that machine learning models substantially improve the out-of-sample performance of stock and bond characteristics in predicting future stock and bond returns. Although both stock and bond characteristics provide strong forecasting power for both stock and bond returns, stock (bond) characteristics do not offer significant incremental predictive power above and beyond bond (stock) characteristics in predicting bond (stock) returns. The results also indicate that stock (bond) characteristics are cash flow (discount rate) predictors and stock (bond) return predictability is driven by mispricing (risk) phenomenon.

2.唐国豪

个人简介:唐国豪,湖南大学金融与统计学院副教授,硕士生导师、金融工程系副主任。中央财经大学金融学博士、美国圣路易斯华盛顿大学访问学者。研究方向为实证资产定价、机器学习、行为金融。主要论文发表于金融学国际顶级期刊Journal of Quantitative and Financial Analysis、Journal of Banking and Finance、金融研究、经济学季刊等国内外高水平期刊。主持国家自然科学基金青年项目、湖南省自然科学基金青年项目。获得校优秀教学奖、院科研突破奖、院教学比赛一等奖、校教学比赛二等奖、本科优秀毕业论文指导老师、指导本科生SIT项目国家级立项等。

题目:基于自编码机器学习的金融大数据资产定价研究

摘要:本文在中国股票市场上,使用改进的自编码机器学习方法和包含近百个公司特征变量的金融大数据,对股票收益进行预测,并对自编码因子的收益预测来源进行全面的宏观经济分析。本文的研究发现,自编码因子能够从包含公司特征的大量信息中提取到有效的收益预测信号,并在横截面上获得显著的超额收益。在对因子重要度的研究中,本文发现我国股票市场异象具有时变特征。此外,本文从宏观经济状态和经济政策两个角度进行分析,研究结果表明,基于自编码的投资模型的有效性与宏观经济息息相关,它能够在市场泡沫成分较大和投机气氛较浓的情况下成功对冲市场风险,且能捕捉到财政政策和货币政策所导致的市场环境的变化。

3.杨学伟

个人简介:杨学伟,南京大学工程管理学院副教授、博士生导师。2006年本科毕业于西安电子科技大学数学系,2011年于南开大学概率论与数理统计专业获得理学博士学位,2012年于香港城市大学商学院经济与金融系从事博士后研究。主要研究兴趣包括金融衍生品创新、投资者行为与风险管理等。论文发表于Journal of Financial Economics、Review of Financial Studies、INFORMS Journal on Computing、Mathematical Finance等期刊。

题目:Delta Hedging and Volatility-Price Elasticity: A Two-Step Approach

摘要:Black-Scholes delta does not minimize variance of hedging risk since there exists a long run negative relationship between implied volatility and underlying price. Existing works have successfully captured the aforementioned long run relationship and improved the hedging performance of Black-Scholes delta. We move one step further by incorporating the short-term properties of the volatility-price relationship via a two-step empirical approach. Specifically, we find that the dependency of minimum variance (MV) hedging ratio on volatility-price elasticity is quite stable and that the volatility-price elasticity exhibits characteristic of mean-reverting. Therefore we first estimate a model which can capture the dependency of hedging ratio on volatility-price elasticity, and then substitute predictions of future volatility-price elasticity into the pre-fixed model to obtain the MV hedging ratio. We test the new approach using the S&P 500 daily option data and show that our approach results in higher hedging gain than related methods appeared in recent works.

4.陈坚

个人简介:陈坚,厦门大学经济学院金融系教授、博导。研究方向包括行为金融、实证资产定价、金融工程等领域。科研成果主要发表于Journal of Financial and Quantitative Analysis, Journal of International Money and Finance, Journal of Futures Markets,金融研究,管理科学学报等国内外优秀期刊。作为项目负责人完成国家自然科学基金项目两项。

题目:Labor Hiring and Stock Returns

摘要:Using a novel data from individual resumes of public firm employees, we propose a net labor hiring at the market and find that its monthly innovation predicts the subsequent market excess return negatively. This predictability is robust after controlling for common return predictors and alternative labor-related variables. Moreover, the superior performance of our measure exists out-of-sample and then delivers sizable economic gains for mean-variance investors in asset allocation. Because the net labor hiring is associated with the labor adjustment cost, its return predictability primarily stems from its ability to forecast future unemployment and cash flows.

5.刘彦初

个人简介:刘彦初,中山大学岭南学院院长助理、金融系副教授、博士生导师。香港中文大学金融工程学博士,博士后,中国科学技术大学理学硕士和理学学士。主要研究兴趣为金融科技与金融工程。在管理科学学报、Operations Research、INFORMS Journal on Computing、European Journal of Operational Research、Journal of Futures Markets、Journal of Economic Dynamics and Control、Insurance: Mathematics and Economics等相关领域国内外主流学术期刊上发表30余篇研究论文。担任中国运筹学会金融工程与金融风险管理分会常务理事、广州数字金融协会供应链金融专委会委员、深圳市金融科技协会湾区国际金融科技实验室特聘专家。

题目:Media-expressed Tone, Option Characteristics, and Stock Return Predictability

摘要:We investigate the informational content of a huge assortment of NASDAQ articles about a joint cross-section of S&P 500 stock return data and related single-stock option data. Splitting the articles into a trading-time and an overnight archive, we distill tone from each of them. We show that media-expressed tone impacts option markets and that both option data and tone predict stock returns. This result is robust to partialing out tone, but depends on whether the tone is from the overnight or from the trading-time archive which differ in terms of their thematic content. Overall, we conclude that the informational content of option data for predicting single-stock returns extends beyond the information summarized in tone and traditional market factors.

6.李剑

个人简介:李剑,浙江工商大学国际商学院助理教授,博士毕业于德国法兰克福大学宏观与货币经济学系,主要研究领域为家庭金融、金融稳定、FDI与老龄化等。论文在European Economic Review等国际期刊发表。

题目:Completing the Market: Generating Shadow CDS Spreads by Machine Learning

摘要:We compared the predictive performance of a series of machine learning and traditional methods for monthly CDS spreads, using firms’ accounting-based, market-based and macroeconomics variables for a time period of 2006 to 2016. We find that ensemble machine learning methods (Bagging, Gradient Boosting and Random Forest) strongly outperform other estimators, and Bagging particularly stands out in terms of accuracy. Traditional credit risk models using OLS techniques have the lowest out-of-sample prediction accuracy. The results suggest that the non-linear machine learning methods, especially the ensemble methods, add considerable value to existent credit risk prediction accuracy and enable CDS shadow pricing for companies missing those securities.

7.李斌

个人简介:李斌,武汉大学经济与管理学院金融系教授、博导、副主任,武汉大学金融科技研究中心主任。主要研究领域为金融科技、投资管理与机器学习。论文发表于Journal of Accounting Research、Journal of Futures Markets、Artificial Intelligence、Journal of Machine Learning Research、管理科学学报、中国工业经济等国内外知名金融会计与计算机类期刊。

题目:Selecting Mutual Funds from the stocks they hold: A machine learning approach

摘要:We select mutual funds in real time by combining individual fund holdings and a large number (94) of stock characteristics to compute fund-level exposures to characteristics on the basis of the stocks they hold. The majority of funds are exposed---both positively and negatively--to approximately 40-50 characteristics. In addition, fund performance is non-linearly related to fund characteristics and their interactions. This feature proves important when we predict fund performance, as machine learning methods such as boosted regression trees (BRTs) significantly outperform standard linear frameworks. Our BRT-generated forecasts encompass the ones generated by the predictors of mutual fund performance that have been proposed in the literature so far.