Neural Realized Measures: Forecast-Supervised Intraday Aggregation for Volatility Forecasting

High-frequency volatility forecasting typically separates measurement from prediction: intraday observations are first compressed into a daily realized measure using a fixed aggregation rule, and the resulting statistic is then passed to a forecasting model. This conventional pipeline therefore treats measure construction as forecast-agnostic preprocessing. We ask whether the aggregation stage itself can instead be learned from the downstream forecasting objective while preserving the statistical structure of established realized measures. We propose Neural Realized Measures (NRM), a primitive-preserving, forecast-supervised aggregation framework that retains the original measure-specific intraday contributions but replaces uniform aggregation with positive, mean-normalized context-dependent weights. Uniform weights exactly recover the corresponding raw realized measure (RRM), so the classical construction is nested as a special case. Because NRM remains an explicit weighted sum of established primitive contributions, it also retains direct bar-level attribution. Across SF, HAR, and MIDAS forecasting specifications, NRM is evaluated against corresponding RRM specifications under controlled comparisons. In Heston simulations, NRM reduces RMSE for both contemporaneous and next-day latent integrated variance across all evaluated primitive--regime combinations. On six cryptocurrencies using five-minute data, NRM improves QLIKE in all 54 paired out-of-sample comparisons, with mean gains of 7.23\%, 8.15\%, and 9.22\% under SF, HAR, and MIDAS, respectively. Validation-selected NRM specifications also achieve the lowest mean QLIKE across the six assets among the evaluated approaches. Ablation and SHAP analyses reveal systematic time-of-day and nonlinear local-context effects in the learned allocation.

1. 16 Apr 2025

   Intraday Volatility Forecasting

   The paper presents a new model for predicting high-frequency intraday conditional discrete return densities and volatility using deep learning, which surpasses empirical nonparametric forecasting rules and Space State Models.

   SSRNFeatured 2×

   3shares
2. 13 Nov 2024

   DeepVol: Volatility Forecasting with Dilated Causal Convolutions

   Volatility Forecasting with Dilated Causal Convolutions: The study introduces DeepVol, a model using Dilated Causal Convolutions, which effectively uses high-frequency data to predict next-day market volatility.

   RePEcFeatured 3×

   27shares
3. 20 Dec 2023

   Jumps and Gold Futures Volatility Prediction

   The article studies the efficiency of the jump component in predicting Chinese gold futures volatility using high-frequency data.

   RePEcFeatured 3×

   17shares
4. 12 Jul 2023

   Crude Oil Volatility Prediction with Structural Regime Switching

   The article introduces a new model using Markov regime switching for better prediction of volatility in the crude oil market, outperforming other high-frequency prediction models.

   RePEc

   15shares
5. 16 Jan 2026

   Realised Volatility Forecasting: Machine Learning via Financial Word Embedding

   A new NLP framework shows that adding news text can improve stock volatility forecasting, particularly during volatile times, when combined with standard models.

   arXiv

   20cites
6. 19 Dec 2025

   Asymptotic Expansions for High-Frequency Option Data

   A new method for analyzing financial data helps test for sudden volatility changes, with evidence from SP500 options indicating significant variation.

   SSRNFeatured 2×

   2cites