學(xué)術(shù)講座：Dynamic Effectiveness of Stock Pricing Factors via Random Forest Models

金融學(xué)院SBF論壇2019年第30講

講座題目：Dynamic Effectiveness of Stock Pricing Factors via Random Forest Models

時(shí)間：2019年10月28日（周一）12:20-13:30

地點(diǎn)：博學(xué)925

主講人：向巨

主講人簡(jiǎn)介：向巨，南方科技大學(xué)金融系助理教授。目前的主要研究方向?yàn)榻鹑谌斯ぶ悄?、智能投資及投顧、量化投資和社會(huì)科學(xué)中的人工智能方法，其論文發(fā)表于國(guó)際知名期刊如Journal of Empirical Finance、 Journal of Financial Econometrics、Computational Economics、Mathematical Social Sciences等，并獲得多項(xiàng)研究基金。向博士曾在哥倫比亞大學(xué)作過(guò)研究，并曾全職工作于美國(guó)、歐洲、中國(guó)的多家金融機(jī)構(gòu)，享有深圳市地方級(jí)領(lǐng)軍人才和孔雀人才稱號(hào)，并持有CFA及FRM證書。他研發(fā)的智能系統(tǒng)在2016年3月Alphago與李世石挑戰(zhàn)賽前，成功預(yù)測(cè)出AlphaGo獲勝并給出五盤比賽的勝率為3.1比 1.9。Email: judexiang@yahoo.com, xiangj@sustech.edu.cn

講座內(nèi)容簡(jiǎn)介：In his paper, we select 17 common stock pricing factors from the aspects of company value, growth potential, quality and technical indicators, and analyze the relationship between factors and future stock returns via random forest (RF) and linear regression (LR) models. We then dynamically construct portfolios according to factors’ changing effectiveness. Portfolio returns from Q1 of 2010 to Q2 of 2018 show that both RF and LR multi-factor models have achieved excess returns relative to the benchmark CSI-500 Index. We also find that nonlinear RF models were significantly better than LR ones in terms of risk-adjusted returns, and short-term RF models are better than the long-term ones. Through the trend analysis of factor effectiveness, we believe that the reason lies in short-term inertia in the A-share market, which is related to the speculative nature of A-share market.