A novel HAR-type realized volatility forecasting model using graph neural network

This study introduces a novel model that uses the convolutional neural network (CNN) architecture to fully utilize information from the heterogeneous autoregressive (HAR) family components across different windows in a two-dimensional image format, to forecast the direction of stock market volatility. The proposed model, Convolutional Neural Network-based Heterogeneous Autoregressive-Kitchen Sink (CNN-HAR-KS) model, leverages the CNN model's automated signal generation capabilities while incorporating the HAR components to ensure interpretability and comparability with traditional volatility forecasting models. Empirical analysis in China's stock market reveals that in terms of forecast accuracy, the CNN-HAR-KS model outperforms alternative models, such as logistic or machine learning models, using the same input variables. We construct investment portfolios based on realized volatility forecasts to further explore the CNN-HAR-KS model's economic value. The CNN-HAR-KS model yields the highest daily Sharpe ratio of 0.043, indicating its ability to generate better risk-adjusted returns than alternative models. Further analysis suggests that the proposed model can be applied to other realized volatility-related classification problems. • “CNN-HAR-KS model” - This is a proposed novel model that combines Convolutional Neural Network (CNN) architecture with the Heterogeneous Autoregressive (HAR) family components to forecast stock market volatility direction. • “Automated signal generation” - The CNN-HAR-KS model leverages the advantages of the CNN model in automating the process of generating signals, making it more efficient and accurate. • “Interpretability and comparability” - The model incorporates HAR components to ensure that its results are interpretable and comparable to traditional volatility forecasting models, providing a balance between advanced technology and established methods. • “Superior forecast accuracy” - Empirical analysis conducted on China's stock market demonstrates that the CNN-HAR-KS model has superior forecast accuracy compared to other models such as Logistic or Machine Learning models when using the same input variables. • “Highest daily Sharpe ratio of 0.043” - The economic value of the CNN-HAR-KS model is further evidenced by its ability to generate superior risk-adjusted returns, indicated by the highest daily Sharpe ratio of 0.043, outperforming alternative models.