CNHARQ: High-Dimensional Multivariate Realized Volatility Forecasts

CNHARQ is a novel framework integrating network-based information to forecast multivariate realized volatilities. The dataset includes files for the Correlation-Based Stochastic Block Model, which reduces parameters from O(N²) to O(NK) by uncovering latent community structures among N assets. Authored by Sixuan He and last updated in 2026, the 436.6 MB collection contains PDF, NPY, CSV, and TXT files.

Use Cases

• Forecasting multivariate realized volatilities based on the CNHARQ model framework.
• Analyzing dynamic community structures in asset volatility co-movement based on the CBSBM method.
• Comparing network-based correlation models against traditional industry classification schemes like GICS for forecasting performance.
• Reducing model dimensionality in high-dimensional financial time series based on the parameter reduction from O(N²) to O(NK).

Strengths

• Includes a novel CNHARQ model framework designed for high-dimensional forecasting.
• The CBSBM method reduces model parameters from O(N²) to O(NK), where K≪N.
• Empirical analysis demonstrates the model captures dynamic community structures and outperforms GICS in out-of-sample forecasting.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect temporal bias inherent to the source platform.

Provenance

Source

figshare

Collection Method

Likely contains empirical financial data analyzed with the proposed CNHARQ and CBSBM models.

Freshness

Last updated 2026-05-11 14:59:59; freshness should be verified.