Paper: arXiv 2408.11739

Authors: Dimitar Kitanovski, Igor Mishkovski, Viktor Stojkoski, Miroslav Mirchev

Abstract

Maintaining a balance between returns and volatility is a common strategy for portfolio diversification, whether investing in traditional equities or digital assets like cryptocurrencies. One approach for diversification is the application of community detection or clustering, using a network representing the relationships between assets. We examine two network representations, one based on a standard distance matrix based on correlation, and another based on mutual information. The Louvain and Affinity propagation algorithms were employed for finding the network communities (clusters) based on annual data. Furthermore, we examine building assets’ co-occurrence networks, where communities are detected for each month throughout a whole year and then the links represent how often assets belong to the same community. Portfolios are then constructed by selecting several assets from each community based on local properties (degree centrality), global properties (closeness centrality), or explained variance (Principal component analysis), with three value ranges (max, med, min), calculated on a maximal spanning tree or a fully connected community sub-graph. We explored these various strategies on data from the S&P 500 and the Top 203 cryptocurrencies with a market cap above 2M USD in the period from Jan 2019 to Sep 2022. Moreover, we study into more details the periods of the beginning of the COVID-19 outbreak and the start of the war in Ukraine. The results confirm some of the previous findings already known for traditional stock markets and provide some further insights, while they reveal an opposing trend in the crypto-assets market.

Complexity vs Empirical Score

  • Math Complexity: 6.5/10
  • Empirical Rigor: 6.0/10 (capped from 8.0: transaction costs not modelled; see penalty flags)
  • Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 8.0)

Why this score: The paper employs advanced mathematical concepts from network science (community detection algorithms, centrality measures, spanning trees) and statistical methods (mutual information, PCA), while also conducting extensive backtests across multiple asset classes (S&P 500 and cryptocurrencies) and specific crisis periods with detailed performance analysis.

Research Flowchart

  flowchart TD
  A["Research Goal: Network-Based Diversification for Stocks vs. Cryptocurrencies"] --> B["Data & Inputs<br>S&P 500 & Top 203 Cryptos<br>Jan 2019 - Sep 2022"]
  B --> C["Network Representation<br>1. Correlation Distance Matrix<br>2. Mutual Information Matrix"]
  C --> D{"Community Detection<br>Algorithms"}
  D --> D1["Louvain"]
  D --> D2["Affinity Propagation"]
  D1 & D2 --> E["Portfolio Construction<br>Selection Strategies:<br>Centrality (Degree/Closeness) & PCA"]
  E --> F["Key Findings/Outcomes"]
  F --> F1["Stocks: Diversification benefits align with traditional finance"]
  F --> F2["Cryptos: Opposing trend observed<br>(High volatility impact)"]
  F --> F3["Network metrics remain viable predictors across asset classes"]

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 network-based portfolio diversification method using correlation and mutual information to build asset networks, detect communities, and select assets via centrality or PCA.

Source sentence
In this paper, we use a network-based approach for portfolio diversification, addressing key issues in traditional investment strategies.

Assumes 4

  • Assume 365 trading days per year for both asset types. market
    Source sentence
    For consistency, we assume 365 trading days for both asset types, and for the missing prices, we assign the previous day’s price.
  • Missing price data are filled with the previous day’s price. data
    Source sentence
    for the missing prices, we assign the previous day’s price.
  • Risk‑free rate is the annualized 13‑week U.S. Treasury bill rate. model
    Source sentence
    where Rf is a risk-free rate, i.e. the annualized 13-week U.S. Treasury bill rate
  • Portfolios are equally weighted, except for market‑cap weighted S&P500 and crypto indices. execution
    Source sentence
    all portfolios constructed within our study are equally weighted, except the S&P500 index and the Top 203 cryptocurrencies that take into account the market capitalization.

Data

  • S&P 500 historical prices · Finnhub market-prices public
    Jan 2019 to Sep 2022 · S&P 500 constituents
    Source sentence
    we collected historical prices for stocks that make up the S&P 500 index using Finnhub…
  • Top 203 cryptocurrencies market-prices public
    Jan 2019 to Sep 2022 · cryptocurrencies with market cap >2M USD
    Source sentence
    the Top 203 cryptocurrencies with a market cap above 2M USD in the period from Jan 2019 to Sep 2022.

Recognised sources: CoinMarketCap

Costs

Not modeled A trading or execution result is reported without costs.

Finds 5

  • In cryptocurrency portfolios, assets with higher centrality tend to exhibit lower volatility, opposite to stock portfolios. performance
    Source sentence
    in cryptocurrency portfolios, assets with higher centrality tend to exhibit lower volatility—a trend that contrasts with the behavior observed in stock portfolios
  • The Louvain method with the correlation co‑occurrence matrix shows the best performance across almost all selection metrics. performance
    Source sentence
    One can argue that the Louvain method with the correlation co-occurrence matrix shows the best performance across almost all selection metrics.
  • Affinity propagation generally yields a better Sharpe ratio than Louvain for cryptocurrency portfolios. performance
    Source sentence
    in general, Affinity propagation yields a better Sharpe ratio than Louvain.
  • Network‑based strategies often outperform baseline portfolios for stocks, but their effectiveness for cryptocurrencies is less clear. performance
    Source sentence
    while many network-based strategies appear to outperform baseline portfolios for
  • Volatility increased during the COVID‑19 pandemic, confirming recent findings. statistical
    Source sentence
    These results corroborate recent findings showing the rising volatility during the pandemic

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 32 related papers). 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.

Cited References (91)

  1. BRCA1 protein products: functional motifs (1996)
  2. The utility of wealth (1952)
  3. Common risk factors in the returns on stocks and bonds (1993)
  4. Equilibrium in a capital asset market (1966)
  5. Security prices, risk, and maximal gains from diversification (1965)
  6. Capital asset prices: A theory of market equilibrium under conditions of risk (1964)
  7. Analysis and assessment of ab initio three-dimensional prediction, secondary structure, and contacts prediction (1999)
  8. Clinical aspects of exhaled nitric oxide
and 83 more
  1. Mesenchymal cells, stromal derived factor-1 and rheumatoid arthritis [abstract] (1999)
  2. Zeolites and synthetic mechanisms (1996)
  3. From prey via endosymbiont to plastids: comparative studies in dinoflagellates (1993)
  4. Innovative oncology (1998)
  5. Proceedings of the First National Conference on Porous Sieves: 27-30 June 1996; Baltimore (1996)
  6. Origin of Eukaryotic Cells (1970)
  7. The alveloar macrophage (1995)
  8. Annual Report (1999)
  9. Wrappers for performance enhancement and obvious decision graphs (1995)
  10. The Mouse Tumor Biology Database — link
  11. Linear and nonlinear market correlations: Characterizing financial crises and portfolio optimization (2017)
  12. Portfolio optimization based on network topology (2019)
  13. Complexity theory and financial regulation (2016)
  14. The price of complexity in financial networks (2016)
  15. Networks in finance (2009)
  16. A tool for filtering information in complex systems (2005)
  17. The structure and function of complex networks (2003)
  18. Structure and dynamics of the global financial network (2016)
  19. Correlation structure and evolution of world stock markets: Evidence from Pearson and partial correlation-based networks (2018)
  20. Networks of equities in financial markets (2004)
  21. Dynamic asset trees and portfolio analysis (2002)
  22. Dynamics of market correlations: Taxonomy and portfolio analysis (2003)
  23. Spread of risk across financial markets: better to invest in the peripheries (2013)
  24. Industry-level media tone and the cross-section of stock returns (2022) — link
  25. A network approach to portfolio selection (2016)
  26. Dynamic portfolio strategy using clustering approach (2017)
  27. Implicit government guarantees and the externality of portfolio diversification: A complex network approach (2021)
  28. A review of two decades of correlations, hierarchies, networks and clustering in financial markets (2021)
  29. Network analysis of global stock markets at the beginning of the coronavirus disease (Covid-19) outbreak (2020)
  30. The evolution of the cryptocurrencies market: A complex networks approach (2020)
  31. Correlation networks among currencies (2006) — link
  32. Collective behavior of cryptocurrency price changes (2018)
  33. Dynamics and Structure of the 30 Largest North American Companies (2009)
  34. Hierarchical structure in financial markets (1999)
  35. Correlation based networks of equity returns sampled at different time horizons (2006) — link
  36. Clustering and hierarchy of financial markets data: advantages of the DBHT (2014)
  37. Exploring graph neural networks for stock market predictions with rolling window analysis (2019) — link
  38. Incorporating corporation relationship via graph convolutional neural networks for stock price prediction (2018)
  39. Network community detection: A review and visual survey (2017) — link
  40. Affinity propagation: clustering data by passing messages (2009)
  41. Community detection in cryptocurrencies with potential applications to portfolio diversification (2021) — link
  42. Cross-correlation dynamics and community structures of cryptocurrencies (2020)
  43. Cryptocurrency Portfolio Diversification Using Network Community Detection (2022)
  44. Portfolio Diversification with Clustering Techniques (2023)
  45. Portfolio optimization through a network approach: Network assortative mixing and portfolio diversification (2024)
  46. Empirical Analyses of Networks in Finance (2018)
  47. Predicting stock market movements using network science: An information theoretic approach (2017)
  48. A network perspective of the stock market (2010)
  49. A network analysis of the Chinese stock market (2009)
  50. Forecasting stock market volatility using (non-linear) Garch models (1996)
  51. Nonlinearity in stock networks (2018)
  52. A comprehensive framework for uncovering non-linearity and Chaos in financial markets: Empirical evidence for four major stock market indices (2020)
  53. Evidence of nonlinearity in daily stock returns (1985)
  54. Nonlinearity in the dynamics of financial markets (2011)
  55. Mutual information: a measure of dependency for nonlinear time series (2004)
  56. Networks in financial markets based on the mutual information rate (2014)
  57. Development of stock correlation networks using mutual information and financial big data (2018)
  58. Introduction to econophysics: correlations and complexity in finance (1999)
  59. Theory of Financial Risk and Derivative Pricing: From Statistical Physics to Risk Management (2003)
  60. Cleaning large correlation matrices: Tools from Random Matrix Theory (2017) — link
  61. Clustering of financial time series with application to index and enhanced index tracking portfolio (2005)
  62. Clustering Indian stock market data for portfolio management (2010)
  63. Cluster analysis for portfolio optimization (2008)
  64. Building Diversified Portfolios that Outperform Out of Sample (2016)
  65. Hierarchical clustering-based asset allocation (2017)
  66. Graph neural networks for asset management (2021)
  67. Deep graph convolutional reinforcement learning for financial portfolio management–DeepPocket (2021)
  68. A Framework to Perform Asset Allocation Based on Partitional Clustering (2020)
  69. The impact of COVID-19 on stock markets (2020)
  70. Contagion and portfolio management in times of COVID-19 (2021)
  71. Lead–lag relationships in foreign exchange markets (2020)
  72. Correlation patterns in foreign exchange markets (2019)
  73. From the pandemic to the Russia–Ukraine crisis: Dynamic behavior of connectedness between financial markets and implications for portfolio management (2024)
  74. Enhancing portfolio management using artificial intelligence: literature review (2024)
  75. Interpreting the Sharpe ratio when excess returns are negative (2004)
  76. The sharpe ratio (1994)
  77. Network models to improve robot advisory portfolios (2022)
  78. Explainable artificial intelligence for crypto asset allocation (2022)
  79. Portfolio optimization based on network centralities: Which centrality is better for asset selection during global crises? (2024)
  80. Network-based asset allocation strategies (2019)
  81. Safe machine learning (2024)
  82. Portfolio Construction: A Network Approach (2023)
  83. NetVIX—A network volatility index of financial markets (2022)

0 of 91 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.