A novel loss function for neural network models exploring stock realized volatility using Wasserstein Distance
Saved in:
| Title: | A novel loss function for neural network models exploring stock realized volatility using Wasserstein Distance |
|---|---|
| Authors: | Gobato Souto, Hugo, Moradi, Amir |
| Contributors: | Academie International School of Business, HAN University of Applied Sciences, International Business, HAN University of Applied Sciences@@@Academie International School of Business@@@Lectoraten |
| Source: | Decision Analytics Journal. |
| Publisher Information: | HAN University of Applied Sciences, 2024. |
| Publication Year: | 2024 |
| Subject Terms: | Topological data analysis, Neural Networks, Neural basis expansion analysis, Exogenous variables, Temporal fusion transformer, Realized volatility forecasting |
| Description: | This study proposes a novel loss function for neural network models that explores the topological structure of stock realized volatility (RV) data by adding Wasserstein Distance (WD). The study shows that the proposed loss statistically significantly improves the forecast accuracy of neural network models for magnitude-dependent error measures, for example, Root Mean Square Error (RMSE) and Mean Absolute Error (MAE), but not necessarily for relative error measures, such as Quasi-likelihood (QLIKE). Additionally, this research provides user-friendly open-source code for researchers and practitioners to implement the proposed loss function efficiently and quickly. |
| Document Type: | article |
| Access URL: | https://surfsharekit.nl/public/594f25ee-7b3a-4132-94df-9d5dbd848ada https://www.sciencedirect.com/science/article/pii/S2772662223002096 |
| Availability: | http://www.hbo-kennisbank.nl/en/page/hborecord.view/?uploadId=sharekit_han:oai:surfsharekit.nl:594f25ee-7b3a-4132-94df-9d5dbd848ada |
| Accession Number: | edshbo.sharekit.han.oai.surfsharekit.nl.594f25ee.7b3a.4132.94df.9d5dbd848ada |
| Database: | HBO Kennisbank |
Nájsť tento článok vo Web of Science | Abstract: | This study proposes a novel loss function for neural network models that explores the topological structure of stock realized volatility (RV) data by adding Wasserstein Distance (WD). The study shows that the proposed loss statistically significantly improves the forecast accuracy of neural network models for magnitude-dependent error measures, for example, Root Mean Square Error (RMSE) and Mean Absolute Error (MAE), but not necessarily for relative error measures, such as Quasi-likelihood (QLIKE). Additionally, this research provides user-friendly open-source code for researchers and practitioners to implement the proposed loss function efficiently and quickly. |
|---|