Paper: arXiv 2308.07944

Authors: Petr Sokerin, Kristian Kuznetsov, Elizaveta Makhneva, Alexey Zaytsev

Abstract

Portfolio management is an essential part of investment decision-making. However, traditional methods often fail to deliver reasonable performance. This problem stems from the inability of these methods to account for the unique characteristics of multivariate time series data from stock markets. We present a two-stage method for constructing an investment portfolio of common stocks. The method involves the generation of time series representations followed by their subsequent clustering. Our approach utilizes features based on Topological Data Analysis (TDA) for the generation of representations, allowing us to elucidate the topological structure within the data. Experimental results show that our proposed system outperforms other methods. This superior performance is consistent over different time frames, suggesting the viability of TDA as a powerful tool for portfolio selection.

Complexity vs Empirical Score

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

Why this score: The paper employs advanced mathematics from Topological Data Analysis (persistence diagrams, landscapes) which significantly increases complexity, while also providing a detailed experimental pipeline with backtesting on multiple time frames and financial metrics.

Research Flowchart

  flowchart TD
  A["Research Goal: Enhance Portfolio Management<br>using Topological Data Analysis"] --> B["Input: Multivariate Time Series Data<br>from Stock Markets"]
  B --> C["Stage 1: Feature Extraction<br>Topological Data Analysis TDA"]
  C --> D["Stage 2: Representation Learning<br>Time Series Clustering"]
  D --> E["Key Outcome: Constructed Investment Portfolio"]
  E --> F["Key Finding: Consistent Outperformance<br>vs. Traditional Methods"]

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 two‑stage method that generates TDA‑based embeddings of stock time series and clusters them to build a portfolio.

Source sentence
... Our approach utilizes features based on Topological Data Analysis (TDA) for the generation of representations

Assumes 4

  • The investor has limited capital and cannot hold hundreds of stocks. execution
    Source sentence
    We assume that our investor has a limited amount of money and can’t build a portfolio of hundreds of stocks.
  • The number of clusters is set to 11, matching the number of economic sectors. model
    Source sentence
    We choose a number of clusters equal to 11 as a number of economic sectors.
  • Embeddings and clustering are performed only on the training data; test data is used solely for backtesting. data
    Source sentence
    For all datasets, the embeddings and clustering were only provided for training step. The test data was used for financial portfolio backtesting and result estimation.
  • Risk minimization is the primary objective for selecting the best clustering method. model
    Source sentence
    The best clustering method for each case according to the risk metric is presented in the results.

Data

  • Standard & Poor's 500 market prices · Standard & Poor's market-prices public
    2012‑2013 train, 2014‑2017 test; 2015‑20 · S&P 500 constituents
    Source sentence
    As data for time series clustering, we use the market prices of the best American stocks from the Standard and Poor 500 index (S&P 500).

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

Source sentence (the authors' words)
Our method is therefore able to achieve lower risks with lower costs.

Finds 4

  • The TDA PI1 embedding achieves the lowest risk among all tested methods. performance
    Source sentence
    The best algorithm according to our main metric, risk, is the TDA method using the PI1 method.
  • For the 2015‑2019 and 2018‑2022 periods, the TDA model has the lowest risk of all baselines. performance
    Source sentence
    In 2015-2019 and 2018-2022, our model has the lowest risk of all the baselines.
  • The TDA model’s Sharpe ratio is several times higher than that of the S&P 500 index. performance
    Source sentence
    It is worth noting that the Sharpe ratio of our model is several times higher than that of the index.
  • The proposed system outperforms other methods consistently across different time frames. performance
    Source sentence
    Experimental results show that our proposed system outperforms other methods. This superior performance is consistent over different time frames

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 22 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.