Paper: arXiv 2507.20039

Authors: Zihan Lin, Haojie Liu, Randall R. Rojas

Abstract

This study proposes a novel portfolio optimization framework that integrates statistical social network analysis with time series forecasting and risk management. Using daily stock data from the S&P 500 (2020-2024), we construct dependency networks via Vector Autoregression (VAR) and Forecast Error Variance Decomposition (FEVD), transforming influence relationships into a cost-based network. Specifically, FEVD breaks down the VAR’s forecast error variance to quantify how much each stock’s shocks contribute to another’s uncertainty information we invert to form influence-based edge weights in our network. By applying the Minimum Spanning Tree (MST) algorithm, we extract the core inter-stock structure and identify central stocks through degree centrality. A dynamic portfolio is constructed using the top-ranked stocks, with capital allocated based on Value at Risk (VaR). To refine stock selection, we incorporate forecasts from ARIMA and Neural Network Autoregressive (NNAR) models. Trading simulations over a one-year period demonstrate that the MST-based strategies outperform a buy-and-hold benchmark, with the tuned NNAR-enhanced strategy achieving a 63.74% return versus 18.00% for the benchmark. Our results highlight the potential of combining network structures, predictive modeling, and risk metrics to improve adaptive financial decision-making.

Complexity vs Empirical Score

  • Math Complexity: 7.0/10
  • Empirical Rigor: 6.0/10 (capped from 8.5: transaction costs not modelled; see penalty flags)
  • Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 8.5)

Why this score: The paper employs advanced econometric and statistical concepts like VAR and FEVD with detailed mathematical formulations, indicating high math complexity. It demonstrates high empirical rigor by using real-world S&P 500 data, implementing a specific backtest with concrete performance metrics (63.74% return vs. benchmark), and detailing data preprocessing and rolling-window simulations.

Research Flowchart

  flowchart TD
  subgraph Input
      A["Research Goal<br>Portfolio Optimization<br>with Network Structure"]
      B["Data: S&P 500 Daily Returns<br>2020-2024"]
  end

  subgraph Methodology
      C["1. Network Construction<br>VAR + FEVD<br>Estimate dependencies"]
      D["2. Core Structure<br>Minimum Spanning Tree MST<br>Reduce noise"]
      E["3. Selection & Forecast<br>Centrality Ranking + NNAR/ARIMA"]
      F["4. Risk Constraint<br>VaR-based Capital Allocation"]
  end

  subgraph Outcomes
      G["Dynamic Portfolio<br>Trading Simulation"]
      H["Key Findings<br>NNAR Strategy: 63.74% Return<br>Benchmark: 18.00%"]
  end

  A --> B
  B --> C
  C --> D
  D --> E
  E --> F
  F --> G
  G --> H

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a novel portfolio optimization framework that combines statistical network analysis, time series forecasting, and risk management.

Source sentence
This study proposes a novel portfolio optimization framework that integrates statistical social network analysis with time series forecasting and risk management.

Assumes 6

  • The VAR model uses a lag order of one for simplicity. model
    Source sentence
    The choice of a lag order of one reflects a preference for simplicity and is often sufficient to capture short-term dependencies in high-frequency financial data.
  • The forecast horizon h is fixed at 10 days. model
    Source sentence
    In this paper, we fix the forecast horizon h at 10 days, aligning with a short-term investment perspective.
  • White‑noise errors have zero mean and constant variance. statistical
    Source sentence
    ut is the vector of white noise error terms, assumed to have zero mean and constant variance.
  • Risk‑free rate is assumed to be zero. market
    Source sentence
    Rf is the risk‑free rate, which we assume to be zero in this analysis.
Show 2 more
  • Only stocks with less than 10% missing or invalid data are retained. data
    Source sentence
    only those stocks with less than 10% missing or invalid data were retained.
  • Portfolio is long‑only; negative weights are set to zero. execution
    Source sentence
    any resulting negative weights are set to zero to prevent short positions under the long-only assumption.

Data

Backtest sample: 2022-06 – 2023-10 · S&P 500 stocks · equities · daily

  • Daily adjusted closing prices for S&P 500 constituents · Yahoo Finance market-prices public
    2020-01-01 to 2024-12-31 · 490 S&P 500 stocks after filtering
    Source sentence
    In this study, we use daily adjusted closing prices for the constituents of the S&P 500 index over the period from January 1, 2020, to December 31, 2024, collected from Yahoo Finance.
  • S&P 500 index (benchmark) · Yahoo Finance benchmark public
    2020-01-01 to 2024-12-31 · S&P 500 index
    Source sentence
    we collect data for the S&P 500 index itself (ticker symbol ^GSPC) to serve as a benchmark.

Recognised sources: Yahoo Finance

Costs

Unclear - sources disagree The platform classifier and the paper-text extraction give different answers, so no value is asserted.

Platform classifier: Not modeled

Paper text extraction: Partial - Fees and slippage erode returns; the Fixed Portfolio incurs minimal costs because it only trades at the outset.

Source sentence (the authors' words)
In a realistic trading environment, every rebalance generates fees and slippage that erode returns; because the Fixed Portfolio requires only an initial allocation (and no further daily trades), its net performance after costs remains significantly above the benchmark.

Finds

Reported performance (MST + AllAgree + VaR): annual return 85.7% · total return 85.7%

Checked by others

No other paper in the archive has been found to support or contradict these findings. Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.