沪深300高频波动率的预测及应用——基于深度学习的方法

Prediction and Application of High Frequency Volatility of CSI 300——Based on Deep Learning Methods

选取沪深300指数5分钟高频数据,以高频价格序列的强记忆性为切入点,构建基于高频价格序列的长短期记忆模型LSTM。基于已实现波动率(RV)理论计算出真实波动率的预测值,并研究了预测波动率在趋势择时策略中的应用。研究发现:基于高频价格序列的LSTM波动率预测模型的预测能力明显优于其他三种模型,充分发挥了长短期记忆模型的优势,经过该波动率改进的趋势择时策略很好控制了投资风险。

Long Short-Term Memory model(LSTM)is constructed by using 5 minutes high-frequency data of the CSI 300 index.The model is based on the Realized Volatility(RV)theory and calculates the prediction value of the true volatility and studies the application of the predicted volatility in the trend timing strategy.The study finds that the prediction ability of the LSTM volatility prediction model based on the high-frequency price series is significantly better than the other three models,giving full play to the advantages of the Long Short-Term memory model.The trend timing strategy improved by the volatility can well control investment risks.