广义双曲线贝叶斯深度网络模型及应用研究

打开文本图片集

中图分类号：0212.8 文献标识码：A文章编号：1006-8228（2025）08-11-05

Generalized Hyperbolic Bayesian Deep Networks：Statistical Model and Applications

Li Jian1，Bi Xiuchun’，Yang Haofeng² （1.Guizhou University ofFinance and Economics，Guiyang，Guizhou 55oo25，China; 2.Risk Management Department，Industrial BankZhengzhou Branch）

Abstract：Toaddressthelimitationsofdeepneuralnetworksas"blackbox"modelsintermsoftheirinabilitytoquantify predictiveuncertaintyasellastheonstraintsimposedbynoalditributionasumptionsinBayesiandepnetworksthispaper proposesanovelBayesiandeepnetworkmodelbasedontheGeneralizedHyperbolicDistribution（GH）.Empiricalvalidation confimtheefectivenessoftheproposedmodel.Experimentalresultsinfinancialmarketsdemonstratesignificantpeformance improvementunderGHdistributionassmptionscomparedtonormaldistributionassmptions，withreducedstructuraluncertainty. Thisstudy contributes a more generalizable framework for nonlinear time series analysis.

Keywords：Times Series Analysis；Bayesian Deep Networks;DeepLearning;Generalized HyperbolicDistributio

0引言

近年来，深度学习（DeepLearning，DL）因良好的模型预测性能，在大数据分析应用中备受关注。（剩余6186字）