Enhancing Financial Risk Management through LSTM and Extreme Value Theory: A High-Frequency Trading Volume Approach

Abstract

This study employs high-frequency trading volume data in the financial sector to apply the Long Short-Term Memory (LSTM) model for purposes of risk management.By incorporating high-frequency data, which includes trading volume information, with the LSTM model, we develop an LSTM-RV dynamic prediction model for realized volatility. Employing a semi-parametric Extreme Value Theory (EVT) approach, we estimate return quantiles to construct the efficiency risk management model. Empirical analysis reveals that the LSTM-RV prediction model markedly improves prediction accuracy compared to the traditional Heterogeneous Autoregressive (HAR) volatility prediction model. Additionally, the LSTM-RV-EVT model outperforms both the conventional model and models that exclude trading volume information.