New DNN Method Quantifies Time Series Prediction Uncertainty

Researchers have developed a new methodology using standard ReLU Deep Neural Networks (DNNs) to predict time series and quantify the uncertainty associated with these predictions. This approach addresses both future variability and estimation variability from training data by constructing a pertinent prediction interval (PPI). The study explores the consistency of DNN estimators with beta-mixing dependent data and demonstrates that a forward bootstrap series maintains beta-mixing properties and the original time series' stationary distribution, which are crucial for enabling the PPI. The proposed method is validated through simulations and real-data analysis, comparing its performance against standard non-parametric methods. AI

IMPACT Introduces a novel method for uncertainty quantification in time series predictions using DNNs, potentially improving reliability in scientific and financial forecasting.

RANK_REASON The cluster contains an academic paper detailing a new methodology for time series prediction using deep neural networks. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv stat.ML →

• alphaXiv
• arXiv
• beta-mixing
• CatalyzeX Code Finder for Papers
• CORE Recommender
• DagsHub
• deep neural network
• forward bootstrap
• Gotit.pub
• Hugging Face
• Influence Flower
• Relu Deep Neural Networks and Linear Finite Elements
• ScienceCast

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →