Quantile deep learning models for multi-step ahead time series prediction

ArXiv ID: 2411.15674 “View on arXiv”

Authors: Unknown

Abstract

Uncertainty quantification is crucial in time series prediction, and quantile regression offers a valuable mechanism for uncertainty quantification which is useful for extreme value forecasting. Although deep learning models have been prominent in multi-step ahead prediction, the development and evaluation of quantile deep learning models have been limited. We present a novel quantile regression deep learning framework for multi-step time series prediction. In this way, we elevate the capabilities of deep learning models by incorporating quantile regression, thus providing a more nuanced understanding of predictive values. We provide an implementation of prominent deep learning models for multi-step ahead time series prediction and evaluate their performance under high volatility and extreme conditions. We include multivariate and univariate modelling, strategies and provide a comparison with conventional deep learning models from the literature. Our models are tested on two cryptocurrencies: Bitcoin and Ethereum, using daily close-price data and selected benchmark time series datasets. The results show that integrating a quantile loss function with deep learning provides additional predictions for selected quantiles without a loss in the prediction accuracy when compared to the literature. Our quantile model has the ability to handle volatility more effectively and provides additional information for decision-making and uncertainty quantification through the use of quantiles when compared to conventional deep learning models.

Keywords: Quantile regression, Multi-step time series prediction, Uncertainty quantification, Deep learning, Volatility forecasting, Cryptocurrencies

Complexity vs Empirical Score

• Math Complexity: 5.5/10
• Empirical Rigor: 7.0/10
• Quadrant: Holy Grail
• Why: The paper utilizes advanced statistical concepts like quantile regression and deep learning architectures (LSTM, CNN) but remains accessible to quant practitioners, while the empirical rigor is high due to implementation on real cryptocurrency datasets (Bitcoin, Ethereum), use of benchmark time series data, and explicit comparison with conventional models.

  flowchart TD
    Start["Research Goal: Develop & Evaluate<br/>Quantile Deep Learning<br/>for Multi-Step Time Series"] --> Inputs["Data & Inputs<br/>Bitcoin, Ethereum, Benchmark Datasets<br/>(High Volatility/Extreme Conditions)"]
    Inputs --> Method["Key Methodology<br/>Quantile Regression Loss Function<br/>Embedded in Deep Learning Models<br/>Multi-Step Ahead Forecasting"]
    Method --> Process["Computational Process<br/>Train Models (Univariate/Multivariate)<br/>Evaluate against Baselines<br/>Quantify Uncertainty"]
    Process --> Results["Key Findings & Outcomes<br/>Improved Extreme Value Handling<br/>Accurate Quantile Predictions<br/>Better Uncertainty Quantification<br/>No Loss in Accuracy"]