Multi-Horizon Realized Variance Forecasting across Asset Classes with a Weighted Ensemble

Q: What does this research mean for the field?

A weighted ensemble combining Ridge and XGBoost achieves superior performance in short-term (1-22 days) and long-term (180-252 days) realized variance forecasting across multiple asset classes, whereas Ridge regression is optimal for medium-term (60-120 days) horizons. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Sang-Hoon Park; Ayeong Seong; Yu Ha; Gun-Woo Kim

doi:10.14801/jkiit.2026.24.8.27

August 31, 2026The Journal of Korean Institute of Information Technology

Multi-Horizon Realized Variance Forecasting across Asset Classes with a Weighted Ensemble

View at Publisher

Ask AI

Bookmark

Share

Key Points

To design and validate an accurate multi-asset, multi-horizon realized variance forecasting system by coupling structured machine-learning features with a weighted ensemble architecture.
Engineered 35 features across four domains (HAR-based, high-frequency proxies, implied volatility surface, and alternative indicators) using daily OHLCV and VIX term structure metrics for 11 equity, bond, and commodity ETFs.
Developed a weighted ensemble (WEns) combining Ridge regression and XGBoost, evaluating it against linear, decision-tree, and neural network baselines across eight horizons (1, 5, 22, 60, 90, 120, 180, and 252 trading days).
The weighted ensemble outperformed all benchmark models over short horizons (1 to 22 trading days) and long horizons (180 to 252 trading days).
Ridge regression achieved superior predictive performance over intermediate horizons spanning 60 to 120 trading days.
Feature importance consensus and sensitivity tests confirmed model stability against maturity mismatches, missing observations, and asset-specific proxy substitutions.

Implication

Comparative analysis demonstrates superior multi-horizon variance prediction across eleven ETFs using a weighted ensemble, highlighting horizon-dependent model advantages.

Abstract

No abstract yet. The journal hasn't published one we can show.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Park et al. (2026) studied this question.

synapsesocial.com/papers/6a9934cf636c6408cfa7c7fe https://doi.org/10.14801/jkiit.2026.24.8.27

View at Publisher

Ask AI

Bookmark

Share

Multi-Horizon Realized Variance Forecasting across Asset Classes with a Weighted Ensemble

Key Points

Implication

Abstract

Discussion

Cite This Study

Multi-Horizon Realized Variance Forecasting across Asset Classes with a Weighted Ensemble

Key Points

Implication

Abstract

Discussion

Cite This Study

Keep reading

More related papers

More from The Journal of Korean Institute of Information Technology

Keep reading

More related papers

More from The Journal of Korean Institute of Information Technology