Host: The Japan Society for Management Information
Name : Annual Conference of Japan Society for Management Information 2017 Spring
Location : [in Japanese]
Date : March 09, 2017 - March 10, 2017
How to model and forecast the volatility, which is the risk of financial assets, is one of the important issues in the financial institution management. Therefore, many prior studies have proposed various models reflected real financial markets. In this study, we attempt to apply the Recurrent Neural Network architecture (Simple RNN, LSTM, GRU) to modeling the volatility clustering and forecasting the future volatility. Using the Recurrent Neural Network architecture, there is a possibility that we could automatically capture structures of the conditional volatility, as ever we have designed manually. In comparison with the GARCH (1,1) model, we analysis a predictability for the conditional volatility.
