统计与决策2026，Vol.42Issue(15)：44-51,8.DOI:10.13546/j.cnki.tjyjc.2026.15.007
分位数因子增广神经网络分位数回归模型的构建及应用
Construction and Application of Quantile Factor-Augmented Quantile Regression Neural Network Model
摘要
Abstract
Macroeconomic forecasting provides vital reference for national macroeconomic regulation and corporate deci-sion-making.Although traditional factor-augmented regression models can effectively resolve forecasting problems in high-di-mensional data,they are difficult to fully depict the complex nonlinear relationships,distribution heterogeneity and heavy-tailed characteristics that are commonly present among economic variables.To this end,the paper proposes a Quantile Factor-Augment-ed Quantile Regression Neural Network Model,which combines the capacity of quantile factor model to describe the common in-formation at different quantiles with the nonlinear modeling capability of neural networks,in order to simultaneously take into ac-count the strong correlations,heavy-tailed distribution and intricate nonlinear relationships in high-dimensional data.Numerical simulation results demonstrate that the proposed model outperforms comparative models across all evaluation metrics under com-plex data conditions such as outliers and heavy-tailed distributions.Further empirical research is conducted by taking unemploy-ment rate forecasting as an example.The results show that the proposed model achieves high accuracy and favorable forecasting stability in both density prediction and point prediction.关键词
分位数因子/非线性/神经网络分位数回归/因子增广回归Key words
quantile factor/nonlinearity/quantile regression neural network/factor-augmented regression分类
管理科学引用本文复制引用
黄玉婷,傅德印..分位数因子增广神经网络分位数回归模型的构建及应用[J].统计与决策,2026,42(15):44-51,8.基金项目
国家社会科学基金资助项目(23BTJ006) （23BTJ006）
兰州财经大学科研项目(Lzufe2026C-003) （Lzufe2026C-003）
兰州财经大学高等教育研究项目(LJY202615) （LJY202615）
