Achieving Superior Volatility Prediction with Deep Learning
We employ Deep Learning to construct an information set to predict the weekly volatility changes from 2010 to 2016 for all stocks in Standard and Poor’s 500 index based on individual stock, market information and treasury note information. Using deep neural network, we train 50000 models and find the binary prediction results and achieved over 70% accuracy for about 500 stocks, month-on- month, for 52 periods from 2011 to 2016. It indicates that deep learning and related machine learning methodologies can be helpful in risk management and in economic stress tests when considering asset allocation.
Prof. Seth H. Huang is an AI researcher/ quantitative trader focusing on developing proprietary methodologies for trading. He is a Taiwanese American and received his PhD in Economics at Cornell University. He is currently also a Director at Shanghai Advanced Institute of Finance, Shanghai Jiao Tong University. His research field focuses on the Deep Learning application in risk management and quantitative trading in global financial assets. Before Shanghai, he lived in Seoul as a finance professor and researcher. He is a partner and adviser for Jumpgate Technologies, a quantitative, big-data-driven hedge fund. He is also a founder of Aris Capital Group, a research and futures trading firm.