Forecasting the Volatility of Specifc Risk for Stocks with LSTM.

The financial market is a complex dynamic system full of noise. Volatility is an important factor in portfolio allocation and risk management. The rapid development of artificial intelligence makes it possible to predict the volatility of the financial market more accurately. In this paper, we attempt to model the volatility of specific risk for stocks. We use GARCH as a baseline model and explore the effectiveness of LSTM. We propose to pack the specific risk factors of all stocks at each timestamp and feed to the LSTM model, so that the implicit connections among stocks will be modeled. From the empirical experiments on BARRA official data, we observe that the performances of LSTM are obviously better than that of GARCH, in terms of both MAE and MSE. [ABSTRACT FROM AUTHOR]