Pith. sign in

REVIEW 4 major objections 5 minor 71 references

Portfolio Optimization under Dynamic Rebalancing via Topological Data Analysis and News Sentiments

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that a topology-based similarity measure, enriched with news sentiment, selects better-diversified assets than correlation or Euclidean distance, and that the resulting dynamically rebalanced portfolios outperform standard

desk verdict A coherent TDA-based portfolio selection pipeline with a plausible novelty, but the central outperformance claim rests on a cherry-picked, unreleased one-year backtest and needs code/data plus multiple-testing control before I'd credit it. read the letter →

arxiv 2607.21170 v1 pith:DUG5J53H submitted 2026-07-23 q-fin.PM

Divyanee Garg This is my paper
classification q-fin.PM MSC 91G1062H30
keywords PortfoliooptimizationTopologicaldataanalysisAgglomerativeclusteringPersistencediagramFinBERTsentimentDynamicrebalancingAssetselectionSparse
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proposes replacing the usual correlation- or Euclidean-based asset similarity with a topological data analysis (TDA) distance computed on a four-dimensional point cloud of technical indicators plus FinBERT news sentiment. Agglomerative clustering on this distance selects a sparse set of stocks, which are then weighted by a mean–variance model with transaction costs, short-selling limits, and dynamic rebalancing. Using S&P 500 data from 2025, the authors report that the TDA-based, sentiment-enriched portfolios consistently beat the correlation/Euclidean variants, the full-universe model, the equally weighted portfolio, the index, and buy-and-hold strategies on return and reward-risk metrics. The paper also reports positive performance during the U.S.–Israel–Iran conflict, arguing the approach adapts quickly to shifting market perception.

What carries the argument

The central object is the TDA-based distance between assets: each asset's multivariate time series is broken into sliding windows, each window gives a point cloud, a Vietoris–Rips filtration produces a persistence diagram (birth–death pairs of connected components and loops), and the average Wasserstein or average persistence-landscape distance between these diagrams defines asset similarity. Agglomerative clustering on this distance, followed by selecting the top-Sharpe assets from each cluster, performs the asset-selection step that feeds into a dynamic rebalancing mean–variance optimization.

What would settle it

Run the identical eight distance measures and five benchmarks on a long out-of-sample window (for example, 2015–2024) using the same pipeline; if the TDA-sentiment portfolio does not consistently beat correlation and Euclidean variants in mean return and Sharpe ratio across at least three non-overlapping periods, the central claim is falsified.

Watch

Extended reading notes

Core claim

The central finding is that measuring asset dissimilarity with persistent homology—specifically the average Wasserstein distance between persistence diagrams or the average persistence landscape distance—captures nonlinear structural differences that correlation and Euclidean distances miss, and that adding a daily FinBERT sentiment score as a fourth feature sharpens this distinction. Across multiple rebalancing frequencies (3, 5, 10 days), in-sample window sizes (63 vs 84 days), and two clustering thresholds, the topologically filtered portfolios achieve higher out-of-sample mean returns, Sharpe ratios, STARR, and cumulative returns than all comparison methods, while keeping portfolios spar

Load-bearing premise

The out-of-sample claims rely on the assumption that expected returns and covariances estimated from only 42 training days, and the lambda selected on 21 validation days, reliably predict performance over the next 3–10 days.

Editorial extensions

If this is right

  • If the claim holds, asset selection becomes a task of comparing topological shape rather than linear co-movement, potentially improving sparse, concentrated portfolios that avoid hidden nonlinear dependence.
  • Sentiment as an additional feature in the similarity measure implies that news flow can drive portfolio rebalancing even when price-based indicators are slow to react.
  • The frequent 3-day rebalancing outperforming 5- and 10-day variants suggests that sentiment's predictive value decays quickly and that an active, high-turnover strategy can overcome its higher transaction costs.
  • The positive conflict-period results imply that topology-enriched, sentiment-aware filtering may offer relative protection in geopolitical stress, though the mechanism is not isolated.
  • The reported cumulative returns, roughly 2.1 times in nine months, would make this a very strong signal if reproducible out-of-sample.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper's design cannot separate the contribution of the distance measure from the contribution of the clustering/selection procedure; a simpler k-nearest-neighbor or threshold-based selection on the same distances might explain much of the gains.
  • Because the in-sample period is only 42 training days and 21 validation days, the out-of-sample comparisons likely reflect short-memory estimation noise; a natural test is whether the same ranking of methods survives on a decade of data.
  • The reported performance suggests a possible hidden dependence on a few high-momentum stocks selected by the Sharpe-ratio filter, which would make the results sensitive to the specific 2025 sample rather than a general property of TDA.
  • Future work could test whether the sentiment feature adds value beyond a simple rolling mean of daily sentiment, or whether the TDA distance is only a roundabout way of picking low-correlation high-momentum assets.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes a two-stage portfolio construction framework: (i) assets are represented by four features (RSI, SO, MACD, and FinBERT-based sentiment), and pairwise similarity is measured by topological distances (average Wasserstein distance and average persistence-landscape distance) or by correlation/Euclidean baselines; (ii) agglomerative clustering is used to select a sparse set of assets, which are then weighted by a dynamic rebalancing mean-variance (DRMV) model with transaction costs, short-selling, and long/short bounds. Out-of-sample results on S&P 500 constituents from 2025 (plus a U.S.-Israel-Iran conflict extension) are reported for 8 distance measures, two confidence levels, and three rebalancing horizons. The paper claims that TDA-based, sentiment-enhanced filtering consistently outperforms correlation/Euclidean filtering and benchmarks such as Naive, Index, full-universe, and buy-and-hold portfolios.

Significance. If the empirical claims held, the paper would offer a useful, practical integration of TDA, news sentiment, and sparse portfolio construction. The mathematical framework is well-defined: the AWD/ALD distances, the sliding-window point-cloud construction, the clustering procedure, and the constrained DRMV model are all specified in enough detail to be reproduced in principle. However, the reported superiority rests on a one-year backtest with many fitted hyperparameters, post-hoc selection of the best-performing configurations, and no code or data release. The contribution is therefore best viewed as a promising methodology whose empirical significance remains unproven.

major comments (4)
  1. [§7, Tables 4–6] The central claim of 'consistent outperformance' is not statistically supported as stated. The empirical protocol considers 8 distance measures × 2 confidence levels (α=0.95/0.99) × 3 rebalancing horizons (T2=3,5,10), i.e., 48 configurations, and the narrative then highlights the best performers (APLS, AWDS) as the proposed framework. Because the same data are used to select and to evaluate these configurations, the reported edge may be a selection artifact. The one-sided t-tests and Sharpe-tests (Appendix D) are applied only to a cherry-picked subset of comparisons, with no correction for multiple testing. The authors should report all 48 configurations, apply a multiple-testing control (e.g., FDR) or pre-register a single primary specification, and provide p-values and confidence intervals rather than only point estimates.
  2. [§6.2–§6.3] The validation-based selection of λ is contaminated by asset selection. At each rolling step, clustering and asset selection use the full in-sample period T1 = D1 + D2 = 63 days (§6.3), while λ is selected by maximizing the Sharpe ratio on the validation segment D2 = 21 days. Thus the validation performance used to choose λ is not independent of the asset-selection mechanism, because the selected assets were already determined using the same D2 observations. This can inflate apparent out-of-sample performance. The authors should either perform asset selection only on the training segment D1, or use a nested cross-validation scheme that keeps validation data out of the clustering/selection step.
  3. [Table 12; Table 15] There are numerical inconsistencies that undermine confidence in the reported results. Table 12 reports ADD = 0.93796 for B&HI (presumably a typo for 0.093796), and Table 15 reports ADD = -0.08672 for (DRMV)AEI under α=0.95, a negative average drawdown, which is impossible by the paper's own definition (average decline from previous peaks). These errors suggest that the tables have not been carefully verified and may reflect underlying data-processing mistakes. All drawdown-related entries should be recomputed and rechecked.
  4. [§6.1 and §8; overall auditability] The empirical evidence is too thin for the strength of the claim. The sample is one year (2025) for the main analysis, and §8 introduces a conflict period from October 2025 to March 2026, yet §6.1 describes data only through December 2025; the source and extension of the conflict-period data should be clarified. More importantly, no code or data are provided, the hyperparameter search space (λ, α, ρ, L, homology dimensions, distance order p) is large, and the extreme cumulative returns (~2.1× over 9 months) are consistent with overfitting to noise. The authors should provide code/data or a full configuration log, and add robustness checks such as block-bootstrap confidence intervals on performance differences or a comparison against randomly selected configurations.
minor comments (5)
  1. [§6.4] The Sharpe ratio is reported on daily returns without annualization or a statement that it is a daily ratio. Since Sharpe ratios are conventionally annualized, this should be clarified.
  2. [Table 7, §7.2] The buy-and-hold benchmarks B&HS and B&HI are said to be based on AWD 'due to its superior out-of-sample performance'; this is a post hoc choice and should be flagged as such or avoided.
  3. [§4.2] The retention mechanism uses a 'next top 10%' threshold that is an additional free parameter not listed in the hyperparameter discussion; its choice should be justified or subjected to sensitivity analysis.
  4. [Appendix D] The t-test is described as a paired test, but the Sharpe test is only loosely referenced to Ledoit and Wolf (2008); the exact implementation (bootstrap scheme, number of resamples, studentization) is not fully specified.
  5. [General] Several minor typos and notation inconsistencies occur, e.g., 'APL' vs. 'APLS/APLI', and the definition of ADD should state its sign convention (positive for a drawdown) explicitly.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the TDA distances and rebalancing framework are defined independently of the reported out-of-sample results; the only post-hoc admission (footnote 14) is a data-snooping concern, not a circular derivation.

full rationale

The claimed derivation chain—TDA distance measures (AWD/ALD) defined from point clouds of standardized technical + sentiment features (Definitions 4.1–4.2), agglomerative clustering (Appendix C), asset selection by in-sample Sharpe, and the DRMV model with λ chosen on a separate validation segment (Section 6.2)—does not reduce to its inputs by construction. The out-of-sample evaluation uses data not used to fit λ, and the distance measures are not defined in terms of the portfolio returns they later predict. The only explicit post-hoc choice is footnote 14: 'The AWD distance measure is adopted due to its superior out-of-sample performance' when constructing the buy-and-hold benchmarks. This is an admitted data-snooping/selection-bias issue for the benchmark comparison, but it is not a circular derivation: the AWD distance itself is defined independently (Eqs. 1–7), and the central TDA-vs-benchmark comparison is not forced by this choice. The paper's strongest claims about outperformance rest on empirical backtests rather than on a parameter-free prediction, but no equation is equivalent to its input, and no load-bearing self-citation or imported uniqueness theorem is present. Therefore no significant circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical or conceptual entities; it reuses existing tools (TDA, FinBERT, mean-variance). The central claim, however, relies on a set of hand-chosen hyperparameters and domain assumptions about the economic meaning of topological summaries and sentiment scores.

free parameters (6)
  • λ (DRMV risk trade-off) = Grid {0.2,0.4,0.6,0.8,1}; chosen on 21-day validation to maximize Sharpe
    The reported out-of-sample performance depends on this in-sample fit; it is a fitted hyper-parameter, not a derived constant.
  • Clustering percentile threshold α = 0.95 and 0.99
    Hand-selected; determines number of clusters and selected stocks. Results reported for both, but the choice is not derived.
  • Selection proportion ρ = 0.10
    Top 10% of assets selected from each cluster; arbitrary.
  • Retention rank threshold = top 10% of SR
    Assets from previous window retained if ranked within next 10%; ad hoc.
  • Sliding window length L = 21 days
    Hand-set; defines point-cloud windows for TDA.
  • Homology dimensions / distance order p = r=0,1; p implicit (likely 2)
    Choice of topological features; standard but hand-set.
assumptions (5)
  • standard math Persistent homology, Wasserstein distance, and persistence landscape definitions (§3)
    These are standard mathematical constructions used as background.
  • domain assumption The 4D point cloud of standardized RSI, SO, MACD, and sentiment has meaningful topological structure for asset similarity (§3, §4.1)
    No evidence is given that topological features of these technical/sentiment coordinates correspond to economically relevant similarity.
  • domain assumption FinBERT sentiment score S_t = P_pos − P_neg averaged over daily headlines is a valid feature for measuring similarity (§2.2)
    The score is asserted without independent validation against market outcomes or human labels.
  • domain assumption Average-linkage agglomerative clustering with a percentile cut-off yields clusters whose top-Sharpe representatives form a diversified portfolio (§4.2, Appendix C)
    The link between cluster structure and diversification benefits is assumed, not derived.
  • ad hoc to paper Retention mechanism: previous-window assets ranked in the next top 10% of SR are retained (§4.2)
    No theoretical basis is given; the threshold is chosen by hand.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Portfolio Optimization under Dynamic Rebalancing via Topological Data Analysis and News Sentiments." pith.science (2026). https://pith.science/paper/DUG5J53H

@misc{pith2026260721170,
  author       = {Pith},
  title        = {Pith review of: Portfolio Optimization under Dynamic Rebalancing via Topological Data Analysis and News Sentiments},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DUG5J53H}},
  note         = {Machine review of arXiv:2607.21170}
}
read the original abstract

Understanding similarity among financial assets is essential for effective portfolio diversification. This paper proposes a novel sentiment-adjusted portfolio optimization framework that integrates Topological Data Analysis (TDA) with technical indicators and FinBERT-based sentiment scores extracted from financial news. A TDA-based distance measure is employed within an agglomerative clustering framework to identify topologically dissimilar assets for portfolio construction. By incorporating sentiment information, the framework captures rapid changes in market perception and investor behavior that are not reflected by technical indicators alone. Unlike conventional correlation and Euclidean distance based approaches, the proposed method characterizes complex nonlinear relationships through topological summaries. To account for the transient nature of market sentiment, a dynamic rolling-window rebalancing strategy with frequent portfolio updates is adopted. A retention mechanism is further introduced to preserve high-quality assets across consecutive rebalancing windows, thereby reducing portfolio turnover and transaction costs. Extensive empirical analysis on S&P 500 constituents demonstrates that the proposed framework consistently outperforms correlation and Euclidean distance based methods, as well as benchmark strategies including Na\"ive, Index, and full-universe portfolios, in terms of returns and reward-risk performance. Furthermore, the framework exhibits strong robustness by delivering positive performance during periods of heightened market uncertainty, such as the U.S.-Israel-Iran conflict.

Figures

Figures reproduced from arXiv: 2607.21170 by the authors.

Figure 1
Figure 1. Flowchart for sentiment score computation from news headlines. Google Gemini was used to refine and enhance the visual presentation of a flowchart developed by the author. The resulting figure was subsequently reviewed, modified, and finalized by the author. to obtain posterior probabilities for positive, negative, and neutral sentiments, denoted by 𝑃pos, 𝑃neg, and 𝑃neu, respec￾tively, and (iv) constructing the sent… view at source ↗
Figure 2
Figure 2. Illustration of agglomerative clustering. Definition 4.1. For 1 ≤ 𝑝 < ∞, the average 𝑝-Wasserstein distance is defined as 𝑑AWD(𝑍𝑖 , 𝑍𝑗 ) = ∑ 𝐾 𝑘=1 𝑤𝑘 𝑊𝑝 (𝑍𝑖𝑘 , 𝑍𝑗𝑘 ), (6) where 𝑊𝑝 (𝑍𝑖𝑘 , 𝑍𝑗𝑘 ) denotes the 𝑝-Wasserstein distance between the PDs of the 𝑘-th sub-series of 𝑍𝑖 and 𝑍𝑗 , re￾spectively. The weights satisfy 𝑤𝑘 > 0 and ∑𝐾 𝑘=1 𝑤𝑘 = 1. This framework preserves both the geometric information captured by pers… view at source ↗
Figure 3
Figure 3. illustrates the overall asset selection framework, from data collection and preprocessing to clustering and final asset selection. 5. Portfolio formation and trading strategy Assets are selected at the end of each in-sample period using the agglomerative clustering framework and distance measures described in Section 4. This selection process reduces the dimensionality of the investment universe and identifies repre… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Rolling-window framework with cross-validation. The in-sample period is divided into a training period 𝐷1 and a validation period 𝐷2 for selecting the parameter 𝜆 in the (DRMV) model. The optimized portfolio is then evaluated over the out-of-sample period with rebalanc…
Figure 5
Figure 5. Figure 5: Comparison of total number of long and short positions across rolling windows for an in-sample period of 63 trading days and an out-of-sample period of 3 trading days under the (DRMV)AWDS (left), the (DRMV)AWDI (center), and the (DRMV)A (right) models. rolling windows.…
Figure 6
Figure 6. Figure 6: Average Jaccard similarity heatmaps of selected stocks across the eight filtering strategies [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 7
Figure 7. Figure 7: Comparison of average Jaccard similarity on selected stocks across the eight filtering strategies corresponding to different rebalancing frequencies, namely 3-day (left), 5-day (center), and 10-day (right), with 3 month in-sample period and 𝛼 = 95%. • B&HS and B&HI : B…
Figure 8
Figure 8. Figure 8: Out-of-sample cumulative returns for all considered models, including the benchmark models B&HS , B&HI , Naïve, and Index, under a 3-month in-sample period with 3-day rebalancing (left), a 4-month in-sample period with 3-day rebalancing (center), and a 3-month in-sampl…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

71 extracted references · 3 linked inside Pith

  1. [1]

    Outperformanceandtracking: Dynamic asset allocation for active and passive portfolio manage- ment

    AliAl-AradiandSebastianJaimungal. Outperformanceandtracking: Dynamic asset allocation for active and passive portfolio manage- ment. Applied Mathematical Finance, 25(3):268–294, 2018

    2018

  2. [2]

    Technical analysis: power tools for active investors

    Gerald Appel. Technical analysis: power tools for active investors. FT Press, 2005

    2005

  3. [3]

    Finbert: Financial sentiment analysis with pre-trained language models.arXiv preprint arXiv:1908.10063, 2019

    Dogu Araci. Finbert: Financial sentiment analysis with pre-trained language models.arXiv preprint arXiv:1908.10063, 2019

    arXiv 1908

  4. [4]

    Topological features of multivariate distributions: Dependency on the covariance matrix

    Lloyd L Aromi, Yuri A Katz, and Josep Vives. Topological features of multivariate distributions: Dependency on the covariance matrix. Communications in Nonlinear Science and Numerical Simulation, 103:105996, 2021

    2021

  5. [5]

    Investor sentiment in the stock market

    Malcolm Baker and Jeffrey Wurgler. Investor sentiment in the stock market. Journal of Economic Perspectives, 21(2):129–151, 2007

    2007

  6. [6]

    Predicting the direction of stock market prices using tree-based classifiers.The North American Journal of Economics and Finance, 47:552–567, 2019

    Suryoday Basak, Saibal Kar, Snehanshu Saha, Luckyson Khaidem, and Sudeepa Roy Dey. Predicting the direction of stock market prices using tree-based classifiers.The North American Journal of Economics and Finance, 47:552–567, 2019

    2019

  7. [7]

    Sparse and stable markowitz portfolios

    Joshua Brodie, Ingrid Daubechies, Christine De Mol, Domenico Giannone, and Ignace Loris. Sparse and stable markowitz portfolios. Proceedings of the National Academy of Sciences, 106(30):12267– 12272, 2009

    2009

  8. [8]

    Statistical topological data analysis using persistence landscapes.The Journal of Machine Learning Research, 16(1):77–102, 2015

    Peter Bubenik et al. Statistical topological data analysis using persistence landscapes.The Journal of Machine Learning Research, 16(1):77–102, 2015

    2015

Show all 71 references
  1. [9]

    Topological pattern recognition for point cloud data

    Gunnar Carlsson. Topological pattern recognition for point cloud data. Acta Numerica, 23:289–368, 2014

    2014

  2. [10]

    Financial market correlation analysis and stock selection applicationbasedontcn-deepclustering

    Yuefeng Cen, Mingxing Luo, Gang Cen, Cheng Zhao, and Zhigang Cheng. Financial market correlation analysis and stock selection applicationbasedontcn-deepclustering. FutureInternet,14(11):331, 2022

    2022

  3. [11]

    The hasty wisdom of the mob: How market sentiment predicts stock market behavior.Expert Systems with Applications, 77:256–263, 2017

    MS Checkley, D Añón Higón, and H Alles. The hasty wisdom of the mob: How market sentiment predicts stock market behavior.Expert Systems with Applications, 77:256–263, 2017

    2017

  4. [12]

    Lipschitzfunctionshavelp-stablepersistence

    David Cohen-Steiner, Herbert Edelsbrunner, John Harer, and Yuriy Mileyko. Lipschitzfunctionshavelp-stablepersistence. Foundations of Computational Mathematics, 10(2):127–139, 2010

    2010

  5. [13]

    Some improved sparse and stable portfolio optimization problems

    Zhifeng Dai and Fenghua Wen. Some improved sparse and stable portfolio optimization problems. Finance Research Letters, 27:46– 52, 2018

    2018

  6. [14]

    Noise trader risk in financial markets.Journal of Political Economy, 98(4):703–738, 1990

    J Bradford De Long, Andrei Shleifer, Lawrence H Summers, and Robert J Waldmann. Noise trader risk in financial markets.Journal of Political Economy, 98(4):703–738, 1990

    1990

  7. [15]

    Adaptingsentiment lexicons to domain-specific social media texts

    ShuyuanDeng,AtishPSinha,andHuiminZhao. Adaptingsentiment lexicons to domain-specific social media texts. Decision Support Systems, 94:65–76, 2017

    2017

  8. [16]

    Bert: Pre-training of deep bidirectional transformers for language understanding

    JacobDevlin,Ming-WeiChang,KentonLee,andKristinaToutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Tech...

    2019

  9. [17]

    A com- prehensive evaluation of constrained mean-expectile portfolios with short selling.Annals of Operations Research, pages 1–39, 2024

    Vrinda Dhingra, Amita Sharma, and Shiv Kumar Gupta. A com- prehensive evaluation of constrained mean-expectile portfolios with short selling.Annals of Operations Research, pages 1–39, 2024

    2024

  10. [18]

    Using copulae to bound the value-at-risk for functions of dependent risks.Finance and Stochastics, 7(2):145–167, 2003

    Paul Embrechts, Andrea Höing, and Alessandro Juri. Using copulae to bound the value-at-risk for functions of dependent risks.Finance and Stochastics, 7(2):145–167, 2003

    2003

  11. [19]

    Enhanced indexing using cumulative prospect theory utility function with ex- pectile risk.Omega, page 103444, 2025

    Divyanee Garg, Ahmad Zaman Khan, and Aparna Mehra. Enhanced indexing using cumulative prospect theory utility function with ex- pectile risk.Omega, page 103444, 2025

    2025

  12. [20]

    Barcodes: the persistent topology of data.Bulletin of the American Mathematical Society, 45(1):61–75, 2008

    Robert Ghrist. Barcodes: the persistent topology of data.Bulletin of the American Mathematical Society, 45(1):61–75, 2008

    2008

  13. [21]

    Topological data analysis of financial time series: Landscapes of crashes.Physica A: Statistical Mechanics and its Applications, 491:820–834, 2018

    Marian Gidea and Yuri Katz. Topological data analysis of financial time series: Landscapes of crashes.Physica A: Statistical Mechanics and its Applications, 491:820–834, 2018

    2018

  14. [22]

    Sparseportfolio selection via topological data analysis based clustering.Quantitative Finance, 25(8):1261–1291, 2025

    AnubhaGoel,DamirFilipović,andPuneetPasricha. Sparseportfolio selection via topological data analysis based clustering.Quantitative Finance, 25(8):1261–1291, 2025

    2025

  15. [23]

    Risk reduced sparseindextrackingportfolio:Atopologicaldataanalysisapproach

    Anubha Goel, Puneet Pasricha, and Juho Kanniainen. Risk reduced sparseindextrackingportfolio:Atopologicaldataanalysisapproach. Omega, page 103432, 2025

    2025

  16. [24]

    Topological data analysis in investment decisions.Expert Systems with Applications, 147:113222, 2020

    Anubha Goel, Puneet Pasricha, and Aparna Mehra. Topological data analysis in investment decisions.Expert Systems with Applications, 147:113222, 2020

    2020

  17. [25]

    Pairs trading via unsupervised learning.European Journal of Operational Research, 307(2):929–947, 2023

    Chulwoo Han, Zhaodong He, and Alenson Jun Wei Toh. Pairs trading via unsupervised learning.European Journal of Operational Research, 307(2):929–947, 2023

    2023

  18. [26]

    Predicting stock prices with finbert- lstm:Integratingnewssentimentanalysis

    WenjunGu,YihaoZhong,ShizunLi,ChangsongWei,LitingDong, Zhuo yue Wang, and Chao Yan. Predicting stock prices with finbert- lstm:Integratingnewssentimentanalysis. In Proceedingsofthe2024 8th International Conference on Cloud and Big Data Computing, pages 67–72, 2024

    2024

  19. [27]

    Clustering stocks using partial correlation coefficients

    Sean S Jung and Woojin Chang. Clustering stocks using partial correlation coefficients. Physica A: Statistical Mechanics and its Applications, 462:410–420, 2016

    2016

  20. [28]

    An efficient hybrid metaheuristic algorithm for cardinality constrained portfolio optimization

    Can B Kalayci, Olcay Polat, and Mehmet A Akbay. An efficient hybrid metaheuristic algorithm for cardinality constrained portfolio optimization. Swarm and Evolutionary Computation, 54:100662, 2020

    2020

  21. [29]

    Time series classification via topological data analysis

    Alperen Karan and Atabey Kaygun. Time series classification via topological data analysis. Expert Systems with Applications, 183:115326, 2021. First Author et al.:Preprint submitted to Elsevier Page 26 of 28 Portfolio Optimization under Dynamic Rebalancing

    2021

  22. [30]

    Keynes as an investor

    John Maynard Keynes, E Johnson, and D Moggridge. Keynes as an investor. The Collected Works of John Maynard Keynes, 12:1–113, 1983

    1983

  23. [31]

    Finllama: Financial sentiment classification for algorithmic trading applications

    Thanos Konstantinidis, Giorgos Iacovides, Mingxue Xu, Tony G Constantinides, and Danilo Mandic. Finllama: Financial sentiment classification for algorithmic trading applications. arXiv preprint arXiv:2403.12285, 2024

    2024 arXiv

  24. [32]

    Sparse index clones via the sorted𝓁1-norm

    Philipp J Kremer, Damian Brzyski, Małgorzata Bogdan, and Sandra Paterlini. Sparse index clones via the sorted𝓁1-norm. Quantitative Finance, 22(2):349–366, 2022

    2022

  25. [33]

    Robustperformancehypothesistest- ing with the sharpe ratio.Journal of Empirical Finance, 15(5):850– 859, 2008

    OliverLedoitandMichaelWolf. Robustperformancehypothesistest- ing with the sharpe ratio.Journal of Empirical Finance, 15(5):850– 859, 2008

    2008

  26. [34]

    Large language models in finance (finllms)

    Jean Lee, Nicholas Stevens, and Soyeon Caren Han. Large language models in finance (finllms). Neural Computing and Applications, 37(30):24853–24867, 2025

    2025

  27. [35]

    Clusteringalgorithmsforrisk-adjusted portfolio construction.Procedia Computer Science, 108:1334–1343, 2017

    DiegoLeón,ArbeyAragón,JavierSandoval,GermánHernández,An- drésArévalo,andJaimeNiño. Clusteringalgorithmsforrisk-adjusted portfolio construction.Procedia Computer Science, 108:1334–1343, 2017

    2017

  28. [36]

    Risk measurement: An introduction to value at risk

    Thomas J Linsmeier and Neil D Pearson. Risk measurement: An introduction to value at risk. 1996

    1996

  29. [37]

    Whenisaliabilitynotaliability? textual analysis, dictionaries, and 10-ks

    TimLoughranandBillMcDonald. Whenisaliabilitynotaliability? textual analysis, dictionaries, and 10-ks. The Journal of Finance, 66(1):35–65, 2011

    2011

  30. [38]

    Extracting insights from the shape of complex data using topology

    Pek Y Lum, Gurjeet Singh, Alan Lehman, Tigran Ishkanov, Mikael Vejdemo-Johansson, Muthu Alagappan, John Carlsson, and Gunnar Carlsson. Extracting insights from the shape of complex data using topology. Scientific Reports, 3(1):1236, 2013

    2013

  31. [39]

    Comparativeadvancesinfinancialsentimentanalysis: A review of bert, finbert, and large language models

    ManishBarathMahendran,AswinKumarGokul,PoornimaLakshmi, andSPavithra. Comparativeadvancesinfinancialsentimentanalysis: A review of bert, finbert, and large language models. In 2025 3rd International Conference on Intelligent Data Communication Technologies and Internet of Thing...

    2025

  32. [40]

    Pairs trading with topological data analysis

    Sourav Majumdar and Arnab Kumar Laha. Pairs trading with topological data analysis. International Journal of Theoretical and Applied Finance, 26(08):2450002, 2023

    2023

  33. [41]

    Sentiment-aware portfolio optimiza- tion: Cvar-based diversification with deep reinforcement learning

    Lamukanyani A Mantshimuli. Sentiment-aware portfolio optimiza- tion: Cvar-based diversification with deep reinforcement learning. IEEE Access, 2025

    2025

  34. [42]

    JournalofFinance ,7(1):71– 91, 1952

    HarryMMarkowitz.Portfolioselection. JournalofFinance ,7(1):71– 91, 1952

    1952

  35. [43]

    Time series clustering for high-dimensional portfolio selection: a comparative study: R

    Raffaele Mattera, Germana Scepi, and Parmjit Kaur. Time series clustering for high-dimensional portfolio selection: a comparative study: R. mattera et al.Soft Computing, 29(8):4219–4231, 2025

    2025

  36. [44]

    Multiscale partial correlation clustering of stock marketreturns

    Antonis A Michis. Multiscale partial correlation clustering of stock marketreturns. JournalofRiskandFinancialManagement ,15(1):24, 2022

    2022

  37. [45]

    Abib- liometric review of portfolio diversification literature.International Review of Financial Analysis, 90:102836, 2023

    MilenaMigliavacca,JohnWGoodell,andAndrea Paltrinieri. Abib- liometric review of portfolio diversification literature.International Review of Financial Analysis, 90:102836, 2023

    2023

  38. [46]

    Constructionofminimum spanning trees from financial returns using rank correlation.Physica A: Statistical Mechanics and its Applications, 566:125605, 2021

    TristanMillingtonandMahesanNiranjan. Constructionofminimum spanning trees from financial returns using rank correlation.Physica A: Statistical Mechanics and its Applications, 566:125605, 2021

    2021

  39. [47]

    Walk- ing down wall street with a tablet: A survey of stock market predic- tions using the web.Journal of Economic Surveys, 30(2):356–369, 2016

    Michela Nardo, Marco Petracco-Giudici, and Minás Naltsidis. Walk- ing down wall street with a tablet: A survey of stock market predic- tions using the web.Journal of Economic Surveys, 30(2):356–369, 2016

    2016

  40. [48]

    Position: Topological deep learning is the new frontier for relational learning.Proceedings of Machine Learning Research, 235:39529, 2024

    Theodore Papamarkou, Tolga Birdal, Michael Bronstein, Gunnar Carlsson,JustinCurry,YueGao,MustafaHajij,RolandKwitt,Pietro Lio, Paolo Di Lorenzo, et al. Position: Topological deep learning is the new frontier for relational learning.Proceedings of Machine Learning Research, 235:...

    2024

  41. [49]

    Clustering and portfolio selection problems: A unified framework

    Justo Puerto, Moisés Rodríguez-Madrena, and Andrea Scozzari. Clustering and portfolio selection problems: A unified framework. Computers & Operations Research, 117:104891, 2020

    2020

  42. [50]

    Comprehensive survey on hierarchical clustering algorithms and the recent developments.Artificial Intelligence Review, 56(8):8219– 8264, 2023

    XingchengRan,YueXi,YonggangLu,XiangwenWang,andZhenyu Lu. Comprehensive survey on hierarchical clustering algorithms and the recent developments.Artificial Intelligence Review, 56(8):8219– 8264, 2023

    2023

  43. [51]

    Intraday online investor sentiment and return patternsintheusstockmarket

    Thomas Renault. Intraday online investor sentiment and return patternsintheusstockmarket. JournalofBanking&Finance ,84:25– 40, 2017

    2017

  44. [52]

    Topo- logical data analysis for portfolio management of cryptocurren- cies

    RodrigoRivera-Castro,PolinaPilyugina,andEvgenyBurnaev. Topo- logical data analysis for portfolio management of cryptocurren- cies. In 2019 International Conference on Data Mining Workshops (ICDMW), pages 238–243. IEEE, 2019

    2019

  45. [53]

    Deviation measures in risk analysis and optimization.University of Florida, Department of Industrial & Systems Engineering Working Paper, (7), 2002

    R Tyrrell Rockafellar, Stanislav P Uryasev, and Michael Zabarankin. Deviation measures in risk analysis and optimization.University of Florida, Department of Industrial & Systems Engineering Working Paper, (7), 2002

    2002

  46. [54]

    Silhouettes: a graphical aid to the interpretation and validation of cluster analysis

    Peter J Rousseeuw. Silhouettes: a graphical aid to the interpretation and validation of cluster analysis. Journal of Computational and Applied Mathematics, 20:53–65, 1987

    1987

  47. [55]

    The Sharpe ratio

    William F Sharpe. The Sharpe ratio. Streetwise–the Best of the Journal of Portfolio Management, pages 169–185, 1998

    1998

  48. [56]

    Stockprice prediction using sentiment analysis

    ThendoSidogi,RendaniMbuvha,andTshilidziMarwala. Stockprice prediction using sentiment analysis. In 2021 IEEE International Conference on Systems, Man, and Cybernetics (SMC), pages 46–51. IEEE, 2021

    2021

  49. [57]

    Divergence of sentiment and stock market trading.Journal of Banking & Finance, 78:130–141, 2017

    Antonios Siganos, Evangelos Vagenas-Nanos, and Patrick Verwi- jmeren. Divergence of sentiment and stock market trading.Journal of Banking & Finance, 78:130–141, 2017

    2017

  50. [58]

    Portfolioselectionviatopologicaldataanalysis

    Petr Sokerin, Kristian Kuznetsov, Elizaveta Makhneva, and Alexey Zaytsev. Portfolioselectionviatopologicaldataanalysis. In Sixteenth International Conference on Machine Vision (ICMV 2023), volume 13072, pages 371–379. SPIE, 2024

    2023

  51. [59]

    Detecting strange attractors in turbulence

    Floris Takens. Detecting strange attractors in turbulence. In Dy- namical Systems and Turbulence, Warwick 1980: proceedings of a symposium held at the University of Warwick 1979/80, pages 366–

    1980

  52. [60]

    Llama: Open and efficient foundationlanguagemodels

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. Llama: Open and efficient foundationlanguagemodels. arXivpreprintarXiv:2302.13971 ,2023

    2023 arXiv

  53. [61]

    Chun Wang, Zhongming Wu, Wei Xu, and Yu Yuan. An exact algorithm for a cardinality-constrained index tracking model consid- ering investment preferences in portfolio optimization.Journal of Industrial and Management Optimization, 22(1):612–641, 2026

    2026

  54. [62]

    Informativeness of the market news sentiment in the taiwan stock market

    Yu-Chen Wei, Yang-Cheng Lu, Jen-Nan Chen, and Yen-Ju Hsu. Informativeness of the market news sentiment in the taiwan stock market. The North American Journal of Economics and Finance, 39:158–181, 2017

    2017

  55. [63]

    Stockmarket one-day ahead movement prediction using disparate data sources

    BinWeng,MohamedAAhmed,andFadelMMegahed. Stockmarket one-day ahead movement prediction using disparate data sources. Expert Systems with Applications, 79:153–163, 2017

    2017

  56. [64]

    European Journal of Operational Research, 319(3):820–833, 2024

    Zhongming Wu, Kexin Sun, Zhili Ge, Zhihua Allen-Zhao, and Tiey- ongZeng.Sparseportfoliooptimizationvia 𝓁1over𝓁2regularization. European Journal of Operational Research, 319(3):820–833, 2024

    2024

  57. [65]

    An efficient global optimal method for cardinality constrained portfolio optimization

    Wei Xu, Jie Tang, Ka Fai Cedric Yiu, and Jian Wen Peng. An efficient global optimal method for cardinality constrained portfolio optimization. INFORMS Journal on Computing, 36(2):690–704, 2024

    2024

  58. [66]

    Dynamic rebalancing portfolio models with analysesofinvestorsentiment

    Jing-Rung Yu, W Paul Chiou, Cing-Hung Hung, Wen-Kuei Dong, and Yi-Hsuan Chang. Dynamic rebalancing portfolio models with analysesofinvestorsentiment. InternationalReviewofEconomics& Finance, 77:1–13, 2022

    2022

  59. [67]

    Computers&IndustrialEngineering ,109:48–58, 2017

    Jing-RungYu,Wan-JiunPaulChiou,andRen-TingLiu.Incorporating transactioncosts,weightingmanagement,andfloatingrequiredreturn inrobustportfolios. Computers&IndustrialEngineering ,109:48–58, 2017

    2017

  60. [68]

    Portfolio rebalancing model us- ing multiple criteria

    Jing-Rung Yu and Wen-Yi Lee. Portfolio rebalancing model us- ing multiple criteria. European Journal of Operational Research, 209(2):166–175, 2011. First Author et al.:Preprint submitted to Elsevier Page 27 of 28 Portfolio Optimization under Dynamic Rebalancing

    2011

  61. [69]

    Dynamic portfolio optimization with the marcos approach under uncertainty

    Pengrui Yu, Zhipeng Ge, Xiaomin Gong, and Xiao Cao. Dynamic portfolio optimization with the marcos approach under uncertainty. International Review of Financial Analysis, 96:103565, 2024

    2024

  62. [70]

    Generalized adjusted rand indices for cluster ensembles

    Shaohong Zhang, Hau-San Wong, and Ying Shen. Generalized adjusted rand indices for cluster ensembles. Pattern Recognition, 45(6):2214–2226, 2012

    2012

  63. [71]

    Bigdataandportfoliooptimization:Anovelapproachintegratingdea with multiple data sources.Omega, 104:102479, 2021

    Zhongbao Zhou, Meng Gao, Helu Xiao, Rui Wang, and Wenbin Liu. Bigdataandportfoliooptimization:Anovelapproachintegratingdea with multiple data sources.Omega, 104:102479, 2021. First Author et al.:Preprint submitted to Elsevier Page 28 of 28

    2021

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.