Directional Risk-Averse Integrated Loss Strategy for Time Series

Qixuan Fu, Xi Chen, Chao Liu, Xiaosong Ding
Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing, PMLR 278:434-439, 2025.

Abstract

This study focuses on time series forecasting for risk-averse decision-makers, emphasizing trend direction over precise numerical predictions. Traditional methods like MSE fail to capture directional accuracy, which is crucial for risk-averse decisions. While techniques like DILATE improve time alignment, they still rely on numerical metrics. We introduce DRAILS (Directional Risk-Averse Integrated Loss Strategy for time series), a novel loss function that prioritizes directional accuracy while maintaining numerical precision. By incorporating a dynamic reward-penalty system inspired by the newsvendor model, DRAILS minimizes directional errors. Our experiments have shown that DRAILS outperforms existing methods in directional accuracy while maintaining competitive numerical results.

Cite this Paper

BibTeX
@InProceedings{pmlr-v278-fu25a, title = {Directional Risk-Averse Integrated Loss Strategy for Time Series}, author = {Fu, Qixuan and Chen, Xi and Liu, Chao and Ding, Xiaosong}, booktitle = {Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing}, pages = {434--439}, year = {2025}, editor = {Zeng, Nianyin and Pachori, Ram Bilas and Wang, Dongshu}, volume = {278}, series = {Proceedings of Machine Learning Research}, month = {25--27 Apr}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v278/main/assets/fu25a/fu25a.pdf}, url = {https://proceedings.mlr.press/v278/fu25a.html}, abstract = {This study focuses on time series forecasting for risk-averse decision-makers, emphasizing trend direction over precise numerical predictions. Traditional methods like MSE fail to capture directional accuracy, which is crucial for risk-averse decisions. While techniques like DILATE improve time alignment, they still rely on numerical metrics. We introduce DRAILS (Directional Risk-Averse Integrated Loss Strategy for time series), a novel loss function that prioritizes directional accuracy while maintaining numerical precision. By incorporating a dynamic reward-penalty system inspired by the newsvendor model, DRAILS minimizes directional errors. Our experiments have shown that DRAILS outperforms existing methods in directional accuracy while maintaining competitive numerical results.} }
Endnote
%0 Conference Paper %T Directional Risk-Averse Integrated Loss Strategy for Time Series %A Qixuan Fu %A Xi Chen %A Chao Liu %A Xiaosong Ding %B Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing %C Proceedings of Machine Learning Research %D 2025 %E Nianyin Zeng %E Ram Bilas Pachori %E Dongshu Wang %F pmlr-v278-fu25a %I PMLR %P 434--439 %U https://proceedings.mlr.press/v278/fu25a.html %V 278 %X This study focuses on time series forecasting for risk-averse decision-makers, emphasizing trend direction over precise numerical predictions. Traditional methods like MSE fail to capture directional accuracy, which is crucial for risk-averse decisions. While techniques like DILATE improve time alignment, they still rely on numerical metrics. We introduce DRAILS (Directional Risk-Averse Integrated Loss Strategy for time series), a novel loss function that prioritizes directional accuracy while maintaining numerical precision. By incorporating a dynamic reward-penalty system inspired by the newsvendor model, DRAILS minimizes directional errors. Our experiments have shown that DRAILS outperforms existing methods in directional accuracy while maintaining competitive numerical results.
APA
Fu, Q., Chen, X., Liu, C. & Ding, X.. (2025). Directional Risk-Averse Integrated Loss Strategy for Time Series. Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing, in Proceedings of Machine Learning Research 278:434-439 Available from https://proceedings.mlr.press/v278/fu25a.html.

Related Material