Skip to content

Large-Scale Asset Selection via Metric Dependence with Enriched High Frequency Information

Home / Papers / Large-Scale Asset Selection via Metric Dependence with Enriched High Frequency Information

Large-Scale Asset Selection via Metric Dependence with Enriched High Frequency Information

Authors: Yangzhou Chen, Shuaida He, Xin Chen Date: 2026-05-04 Paper ID: openalex:2605.02326

Summary

This paper addresses the sensitivity of large-scale portfolio allocation to estimation error by introducing Metric Dependence Screening (MDS). MDS represents asset data as object-valued point-curves, capturing both daily returns and intraday risk dynamics to improve asset selection. By employing a Fréchet variation-based dependence score, the method effectively filters high-dimensional investable universes before standard mean-variance allocation. Theoretical consistency and empirical performance are validated across simulations and a large-scale dataset of Chinese A-share stocks.

Key Contributions

  • Introduces Metric Dependence Screening (MDS), an asset selection procedure leveraging point-curve objects that combine daily returns with intraday risk curves.
  • Establishes theoretical guarantees including concentration, sure selection, and rank consistency under ultrahigh dimensionality and α-mixing time series dependence.
  • Demonstrates improved out-of-sample portfolio performance over return-based and scalar-based benchmarks on a dataset of 2,938 Chinese A-share stocks.

Open Questions & Future Work

Key Concepts

  • metric-dependence-screening: A high-frequency data-driven screening procedure that uses Fréchet variation-based dependence scores to rank assets based on reward and intraday risk dynamics.

Archivist Review

The paper introduces a robust framework for high-dimensional asset selection using object-valued representations. Metric Dependence Screening is approved for its distinct methodological contribution to financial econometrics. One open question regarding market frictions is approved, as it addresses a core limitation in translating theoretical performance gains to practical feasibility, while the other question was rejected as a standard experimental extension.

Approved Concepts

  • Metric Dependence Screening: It is the core methodological contribution of the paper for asset selection in large-scale portfolios.

Approved Open Questions

  • Integrating Market Frictions in Screening: Without accounting for transaction costs and turnover, the reported portfolio gains may be overstated, and the selection method might be practically infeasible for high-frequency or high-turnover rebalancing strategies.

Rejected Candidates

  • [open_question] Alternative High-Frequency Risk Signatures (alternative-high-frequency-risk-signatures) - low_impact: This is a standard research extension suggestion rather than a substantial theoretical or architectural bottleneck.

Metadata & Links

url
https://arxiv.org/abs/2605.02326
paper_id
2605.02326
paper_source
openalex
domain
time-series
tags
time-series-forecastingfinancial-econometricsportfolio-optimization
architectures
datasets
concept_slugs
metric-dependence-screening
dataset_slugs
skill
TimeSeriesSkill
created_at
2026-05-07T06:05:23Z