From Headlines to Holdings: Deep Learning for Smarter Portfolio Decisions

Jiawei Lou; Jinghe Zhang; Yun Lin

REVIEW 5 major objections 5 minor 14 references

From Headlines to Holdings: Deep Learning for Smarter Portfolio Decisions

T0 review · 5 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read An end-to-end LSTM-GAT network that directly optimizes the Sharpe ratio, using price, graph, and news-sentiment features, beats equal-weight and CAPM-MVO benchmarks out-of-sample.

desk verdict A genuinely new end-to-end LSTM-GAT sentiment portfolio model, clearly written, but the empirical headline is not yet credible because of an undisclosed universe-selection step and a frictionless daily-rebalance benchmark; still worth refereeing. read the letter →

arxiv 2509.24144 v2 pith:F3G3GP55 submitted 2025-09-29 q-fin.PM q-fin.CPq-fin.STstat.ML

Yun Lin , Jiawei Lou , Jinghe Zhang This is my paper · ORCID

classification q-fin.PMq-fin.CPq-fin.STstat.ML MSC 91G1068T07

keywords PortfoliooptimizationAssetallocationGraphneuralnetworksattentionLongshort-termmemoryFinancialnewssentimentSharperatioEnd-to-endlearning

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

This paper claims that an end-to-end deep network can replace the traditional forecast-then-optimize pipeline in portfolio management. The network combines an LSTM for each stock's price and sentiment history, a graph attention layer that lets stocks share information through correlations and sector ties, and a final layer that outputs daily portfolio weights; it is trained directly to maximize the Sharpe ratio. On a nine-stock universe that includes the April 2025 tariff shock in the out-of-sample period, all five model variants beat equal-weight and CAPM-based mean-variance benchmarks, with the sentiment-augmented version reaching a 31.23% annualized return and 1.15 Sharpe ratio. The authors read this as evidence that integrating temporal, relational, and sentiment signals in a single objective yields more robust risk-adjusted returns than two-stage methods. They caution that results are upper-bound estimates because trading is assumed frictionless.

What carries the argument

The load-bearing object is the end-to-end weight generator: a shared LSTM encodes 30-day feature windows for each stock, a Graph Attention Network refines those embeddings using a graph whose edges encode sector membership and correlations of returns and sentiment (static or updated every five trading days), and a tanh linear layer produces raw weights that are normalized to sum to one. The entire system is trained with Adam to minimize the negative Sharpe ratio of the resulting portfolio, so the model directly optimizes the investor's risk-adjusted objective rather than a proxy forecast. The graph construction and the Sharpe loss are what distinguish this from a simple price predictor.

What would settle it

Re-run the same models on the same nine stocks with a per-trade transaction cost of 10 basis points plus slippage and compare net-of-cost Sharpe ratios to the equal-weight and CAPM-MVO baselines; if the LSTM-GAT variants no longer dominate, the paper's central claim is empirically falsified. A second decisive check is to repeat the experiment on a larger, independently selected universe with the stock selection made before the test window begins.

Watch

Extended reading notes

Core claim

The central claim is that an LSTM-GAT network trained end-to-end to maximize the Sharpe ratio—rather than first predicting returns and then optimizing—can learn daily portfolio weights that outperform equal-weight and CAPM-MVO benchmarks. The best configuration, which adds sentiment features to price and volume inputs, achieves a 31.23% annualized return and a 1.15 Sharpe ratio out-of-sample versus 20.85%/0.83 and 18.58%/0.84 for the baselines. The authors attribute the gains to three design choices: direct Sharpe optimization avoids compounding forecast errors; the GAT captures evolving cross-asset dependencies; and news sentiment carries information not fully reflected in prices. They also

Load-bearing premise

The load-bearing premise is that the backtest reflects frictionless trading at official prices with no costs, slippage, or market impact, and that the nine-stock universe was selected without peeking at test-window data; if either fails, the reported edge could vanish.

Editorial extensions

If this is right

If the central claim is correct, daily allocations can be produced by a single network trained on portfolio-level objectives, bypassing the unstable two-step forecast-then-optimize loop.
Adding financial-news sentiment to price-based features improved the annualized return and Sharpe ratio by nearly 50% and 39% over the equal-weight benchmark in the tested universe, suggesting news tone carries information not already in prices.
Switching from a static to a weekly-updated correlation graph reduced volatility and drawdown at a small cost in raw return, indicating that adaptive relational structure helps in changing markets.
PCA-based feature compression produced the smallest maximum drawdown during the April 2025 tariff shock, hinting that noise reduction improves stress resilience.
Because the backtest assumes zero transaction costs and daily rebalancing, the reported outperformance is an upper bound; under realistic frictions the gap over baselines would narrow.

Reading between the lines

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

A decisive next test is to add realistic trading costs (e.g., 5–10 basis points per trade) and slippage; if the sentiment variant's Sharpe ratio falls below the benchmarks, the practical claim collapses even if the frictionless claim stands.
The nine stocks were chosen from 50 S&P 500 names using low-correlation and news-coverage filters; since the paper does not state whether this selection was made before or after seeing the test window, a replication with a pre-registered universe is needed to rule out selection bias.
The third-party sentiment scores are black-box; replacing them with an openly documented sentiment model (e.g., a fine-tuned financial transformer) would identify how much of the reported gain depends on the specific sentiment source.
Because the graph has only nine nodes, the GAT's relational learning is severely constrained; extending to a larger universe is the natural stress test for whether relational information, rather than the Sharpe objective alone, drives the gains.

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.

Desk Editor's Note

read the letter

Here's my take on arXiv:2509.24144. The integration is new: an end-to-end LSTM-GAT that outputs portfolio weights directly, trained on the Sharpe objective, with news sentiment features. The authors test it on nine large-cap US stocks over a window that includes the April 2025 tariff shock, and the best variant (v3) beats equal-weight and CAPM-MVO on return and Sharpe. The methodology is described well enough that the work is understandable even without code. The paper also does several things right: the limitation section is candid, calling the results upper-bound and noting the sparse graph, opaque sentiment API, static sectors, and heuristic threshold. The ablation design is sensible—v1 price-only, v2 adds features, v3 adds sentiment, v4 dynamic graph, v5 PCA—and the finding that sentiment helps and dynamic graphs reduce volatility is coherent.

The soft spots are real and, in one case, more serious than the reader's note suggested. The biggest issue is the universe selection. Section 3.1 says nine stocks were chosen from a pool of fifty for low pairwise correlations and sector diversity, but it never says when or on what data the selection was made. If the correlations were computed over the full 2021–2025 sample, the test window is effectively in-sample for the universe, which can inflate the Sharpe of any dynamic strategy. That is lookahead, not a friction issue, and the authors' upper-bound caveat does not address it. The second issue is the uneven benchmark comparison: the model rebalances daily with zero costs while the CAPM-MVO rebalances every 21 days. Realistic frictions could erase much of the reported advantage. The third is statistical: the evaluation rests on a single seed and a single test window with no error bars or significance testing.

None of these is individually fatal, and the central architecture story holds up. But together they mean the headline outperformance is not established beyond this specific backtest. The fixes are straightforward: disclose the universe-selection timing, add multi-seed or bootstrap intervals, include a cost scenario, and consider a non-deep baseline using the same features. I'd want those before citing the results, but the paper deserves a serious referee—it is clearly written, the integration is novel, and the flaws are addressable. I'd bring it to a reading group to discuss the design and the lookahead trap.

Referee Report

5 major / 5 minor

Summary. The paper proposes an end-to-end deep-learning pipeline for daily portfolio allocation that combines LSTM temporal encoding, a Graph Attention Network over stock relationships, and news-sentiment features, with portfolio weights trained directly by maximizing the Sharpe ratio. Five model variants (v1–v5) are evaluated on a fixed nine-stock U.S. universe over an out-of-sample period spanning early 2024 to mid-2025, including the April 2025 tariff shock, and compared against equal-weight and CAPM-MVO benchmarks. The authors report that all LSTM-GAT variants beat both benchmarks on cumulative/annualized return and Sharpe ratio, with the sentiment-augmented v3 model achieving the best results (31.23% annualized return and 1.15 Sharpe ratio vs. 20.85%/0.83 for equal-weight and 18.58%/0.84 for CAPM-MVO). The paper frames the results as a proof of concept and repeatedly acknowledges the zero-transaction-cost, small-universe, and API-sentiment limitations.

Significance. If the results hold, the paper's contribution is a coherent architecture that unifies temporal modeling, relational graph structure, and sentiment in a single differentiable portfolio objective, avoiding the error compounding of two-step prediction-then-optimization pipelines. The inclusion of the April 2025 stress period is a strength, and the use of external benchmarks and a held-out test window is methodologically sound in spirit. The reported tables are internally consistent, and the authors are candid about several limiting assumptions. However, the evidence supporting the central claim is currently fragile: the universe-selection procedure is not time-stamped and may leak test-period information; the evaluation rests on one random seed and one test window; and the zero-cost daily-rebalancing assumption is acknowledged but never stress-tested. These issues are fixable but require additional experiments.

major comments (5)

[Section 3.1] The stock selection procedure is not time-stamped. The nine stocks are chosen from an initial pool of 50 based on 'low pairwise return correlations,' sector diversity, and news coverage, but the paper never states the time window or data period used to compute those correlations. If the correlation criterion was evaluated over the full 2021–2025 sample, then the test window (early 2024–mid 2025) has been used to construct the universe, which is a lookahead bias that can mechanically inflate the out-of-sample performance of any allocation strategy. This is not covered by the 'upper-bound' caveat in Section 3.3, which concerns frictions rather than test-set validity. Please disclose when and on what data the selection was made, and ideally re-run the selection using only pre-test information or show robustness to alternative universe-construction windows.
[Section 3.5 / Table 2] All reported results come from a single random seed (42) and a single out-of-sample window, with no error bars, confidence intervals, or significance tests. The performance gaps in Table 2—for example, the Sharpe ratio difference between Model v1 (0.91) and Model v2 (0.95), or between Model v4 (1.06) and Model v3 (1.15)—could easily be within the noise of one initialization and one market episode. The paper's central claim that all LSTM-GAT variants outperform the benchmarks needs statistical support: report results over multiple seeds (e.g., 5–10) and, if feasible, multiple test windows or a bootstrap of the test-period return series.
[Section 3.3 / Section 4.1] The evaluation assumes zero transaction costs, zero slippage, and daily rebalancing, which the authors themselves describe as producing upper-bound estimates. This assumption is load-bearing because the reported advantage over the equal-weight benchmark is about 10 percentage points of annualized return, and daily rebalancing across nine stocks would incur nontrivial costs in practice. Please add a transaction-cost sensitivity analysis: for example, show how the cumulative-return and Sharpe-ratio gaps change under round-trip costs of 5, 10, and 25 basis points, and report the break-even cost at which the best LSTM-GAT variant no longer beats the benchmarks. Without this, the practical claim of 'smarter portfolio decisions' is not established.
[Section 2.4 / Section 3.5] The Sharpe-ratio loss in Eq. (1) involves a covariance matrix Σ, but the paper never specifies how Σ is computed during training and evaluation. Is it the empirical covariance of the batch's realized returns? A rolling window estimate? A shrinkaged estimator? Is it recomputed at test time? This is essential for reproducibility and for understanding whether the end-to-end gradient flow is well-defined. Please provide the exact definition and estimation procedure for Σ.
[Section 3.4 / Table 8] The comparison across Model v1–v5 is presented as an ablation of features, graph type, and dimensionality reduction, but each variant is independently tuned with Optuna over 50 trials. Consequently, differences between variants (e.g., v3 vs. v4 for static vs. dynamic graphs, or v4 vs. v5 for PCA) are confounded with differences in hyperparameters: Table 8 shows, for instance, that v3 uses LSTM hidden size 32 and 2 layers, while v4 uses 80/1 and v5 uses 32/1. Attributing performance differences to sentiment or dynamic graphs is only valid if the architectural component is varied while holding the optimization procedure and hyperparameter search fixed. Please either run controlled ablations with a shared hyperparameter budget or soften the causal language in Section 4.1.

minor comments (5)

[Section 3.1 / Table 1] The text in Section 3.4 says v3 adds only 'sentiment variance and weighted sentiment,' but Table 1 shows that v3 also adds 'News Count' and 'Average Sentiment.' Please reconcile the description with the table.
[Table 8] The hyperparameter 'gat alpha' for Model v5 is listed as 0.35, but the Optuna search space in Table 7 specifies a range of 0.05–0.30. This inconsistency should be corrected.
[Table 2] The note says 'Values highlighted in blue represent the best performance,' but the manuscript as rendered has no color. Please indicate the best values by boldface or another unambiguous marker.
[Section 3.2] The phrase 'Building the future set required collecting both price and news data' appears to be a typo; presumably 'feature set' was intended.
[Section 5] The limitations section acknowledges the fixed nine-stock universe, zero transaction costs, static GICS sectors, and opaque third-party sentiment APIs. These are appropriate, but the paper does not mention the lack of code/data availability; adding a reproducibility statement would strengthen the contribution.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the LSTM-GAT variants are trained on training-window data and evaluated on a held-out test period against external benchmarks; no fitted quantity is renamed as a prediction.

full rationale

The paper's central claim is that end-to-end LSTM-GAT variants, trained by directly maximizing the training-window Sharpe ratio (Eq. 1), outperform equal-weight and CAPM-MVO benchmarks out of sample. The evaluation uses a held-out final 30% (Section 3.5) that includes the April 2025 tariff shock, and the benchmarks are external strategies (Section 7.4). No parameter used in the result is fitted to the test window: the static graph is built from the training-period correlation matrix (Section 2.3), hyperparameters are chosen via Optuna on the training/validation split, and the reported Sharpe and return metrics are computed from the trained model's allocations on the test period. There is no load-bearing self-citation; references to LSTM, GAT, and prior portfolio papers are standard external methods. The only notable concern is the undisclosed timing of the 50-to-9 stock universe selection in Section 3.1, which could indicate data snooping, but the paper provides no equation or fitted value that makes the test result equivalent to its inputs by construction, so this is a robustness/transparency risk rather than a circular step under the review rules.

Assumptions & free parameters 8 free parameters · 6 assumptions · 0 invented entities

The central claim rests on standard machine-learning machinery taken as given (backprop, Adam, LSTM gating, GAT attention), plus domain assumptions the paper partly flags: the 2021-2023 training period predicts the 2024-2025 test period, AlphaVantage prices are complete and correctly timestamped, MarketAux sentiment scores are meaningful, GICS sectors are static, and CAPM is an admissible benchmark model. The headline numbers also depend on at least eight hand-chosen or validation-fitted constants, most importantly the per-variant Optuna hyperparameters and the 0.5 dynamic-graph threshold, plus an unspecified covariance estimator for Sigma in the Sharpe loss (Eq. 1). No invented entities are introduced. These counts measure what the paper borrows from assumption versus what it contributes.

free parameters (8)

Per-variant hyperparameters (14 Optuna-tuned) = Table 8, e.g., v3: lstm_hidden=32, lstm_layers=2, gat_hidden=64, lr=1.27e-03, batch=32
Each of the five variants was tuned over 50 Optuna trials to maximize validation Sharpe (Section 3.5); the selected configurations are fit to the data and shape the reported test performance.
Dynamic graph correlation threshold = 0.5
Binary edges when absolute 5-day return or sentiment correlation exceeds 0.5 (Section 2.3); a hand-picked heuristic the authors flag as potentially tunable in Section 5.
Feature lookback window r = 30 trading days
Set to balance medium-term dynamics and compute (Section 2.5).
Dynamic graph refresh period = every 5 trading days
Chosen to balance responsiveness and computational efficiency (Sections 2.3, 3.5).
PCA components (Model v5) = 6 of 12 features
Dimensionality reduction for v5; the selection criterion is qualitative ('stability during high-volatility periods', Section 3.4).
Covariance estimator Sigma in Sharpe loss = unspecified
Eq. 1 requires w^T Sigma w; no estimation window or method for Sigma is given anywhere in the paper, so the loss function is underspecified.
Random seed = 42
Single seed across components; point estimates are seed-dependent ('results may vary slightly across different hardware', Section 3.5).
CAPM-MVO baseline design = 252-day beta window, 21-day rebalancing, GMV fallback
Baseline parameters shape the comparison; the MVO rebalances monthly while the DL models rebalance daily at zero cost (Appendix 7.4).

assumptions (6)

standard math Standard neural-network training machinery (backpropagation, Adam, LSTM gates, GAT attention) behaves as specified
Invoked throughout Section 2; unproved background from Hochreiter and Schmidhuber 1997, Velickovic et al. 2018, and Kingma and Ba 2014.
domain assumption The mapping learned on 2021-2023 data generalizes to the 2024-2025 test window
The train/validation/test split in Section 3.5 assumes the learned mapping from features to Sharpe-maximizing weights transfers across time.
domain assumption MarketAux sentiment scores accurately reflect market psychology
Sentiment features (news count, average/variance/weighted sentiment, Table 5) come from a third-party API with undisclosed methods; the paper itself flags this in Section 5.
domain assumption AlphaVantage price data are complete, correctly timestamped, and free of missing values
Section 3.3 states all trading data are 'treated as correctly timestamped and free of missing values.'
domain assumption Static GICS sector membership is a valid basis for graph edges over the whole sample
Static sector edges are used in Section 2.3; the paper acknowledges static sector classification may introduce structural inaccuracies (Section 5).
domain assumption CAPM beta estimation on rolling 252-day windows is a valid expected-return model for the benchmark
Appendix 7.4 estimates expected returns via CAPM beta regressions on rolling windows; standard textbook assumption for the baseline.

how reviews work

0 comments

Cite this review

Pith. "Pith review of From Headlines to Holdings: Deep Learning for Smarter Portfolio Decisions." pith.science (2026). https://pith.science/paper/F3G3GP55

@misc{pith2026250924144,
  author       = {Pith},
  title        = {Pith review of: From Headlines to Holdings: Deep Learning for Smarter Portfolio Decisions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F3G3GP55}},
  note         = {Machine review of arXiv:2509.24144}
}

read the original abstract

Deep learning offers new tools for portfolio optimization. We present an end-to-end framework that directly learns portfolio weights by combining Long Short-Term Memory (LSTM) networks to model temporal patterns, Graph Attention Networks (GAT) to capture evolving inter-stock relationships, and sentiment analysis of financial news to reflect market psychology. Unlike prior approaches, our model unifies these elements in a single pipeline that produces daily allocations. It avoids the traditional two-step process of forecasting asset returns and then applying mean--variance optimization (MVO), a sequence that can introduce instability. We evaluate the framework on nine U.S. stocks spanning six sectors, chosen to balance sector diversity and news coverage. In this setting, the model delivers higher cumulative returns and Sharpe ratios than equal-weighted and CAPM-based MVO benchmarks. Although the stock universe is limited, the results underscore the value of integrating price, relational, and sentiment signals for portfolio management and suggest promising directions for scaling the approach to larger, more diverse asset sets.

Figures

Figures reproduced from arXiv: 2509.24144 by the authors.

**Figure 2.** The split interval of train, validation, and test set [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗

**Figure 3.** Portfolio cumulative return comparison over the test period. [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗

Figures from the paper (6 more)

**Figure 4.** Figure 4: Cumulative excess return relative to the equal-weight portfolio (test period). [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]

**Figure 6.** Figure 6: Model v2 7.7 Model Predicted Weights [PITH_FULL_IMAGE:figures/full_fig_p020_6.png]

**Figure 5.** Figure 5: Model v1 20 [PITH_FULL_IMAGE:figures/full_fig_p020_5.png]

**Figure 7.** Figure 7: Model v3 [PITH_FULL_IMAGE:figures/full_fig_p021_7.png]

**Figure 8.** Figure 8: Model v4 21 [PITH_FULL_IMAGE:figures/full_fig_p021_8.png]

**Figure 9.** Figure 9: Model v5 22 [PITH_FULL_IMAGE:figures/full_fig_p022_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

14 extracted references · 3 canonical work pages

[1]

Accessed: 2025-07-28

Alpha Vantage Inc.Alpha Vantage API.https://www.alphavantage.co/. Accessed: 2025-07-28. 2025

2025
[2]

High-performance stock index trading: making effective use of a deep long short-term memory neural network

Chariton Chalvatzis and Dimitrios Hristu-Varsakelis. “High-performance stock index trading: making effective use of a deep long short-term memory neural network”. In:arXivabs/1902.03125 (May 2019).url:https://arxiv.org/abs/1902.03125

arXiv 1902
[3]

Graph neural networks for deep portfolio optimization

¨Omer Ekmekc ¸io˘glu and Mustafa C ¸ . Pınar. “Graph neural networks for deep portfolio optimization”. In:Neural Computing and Applications35 (2023), pp. 20663–20674.doi:10.1007/s00521-023- 08862-w.url:https://doi.org/10.1007/s00521-023-08862-w

work page doi:10.1007/s00521-023- 2023
[4]

Louis.3-Month Treasury Bill: Secondary Market Rate (TB3MS).https: //fred.stlouisfed.org/series/TB3MS

Federal Reserve Bank of St. Louis.3-Month Treasury Bill: Secondary Market Rate (TB3MS).https: //fred.stlouisfed.org/series/TB3MS. Accessed: 2025-07-28. 2025

2025
[5]

Long Short-Term Memory

Sepp Hochreiter and J¨ urgen Schmidhuber. “Long Short-Term Memory”. In:Neural Computation9.8 (Nov. 1997), pp. 1735–1780.doi:10.1162/neco.1997.9.8.1735.url:https://doi.org/10. 1162/neco.1997.9.8.1735

work page doi:10.1162/neco.1997.9.8.1735.url:https://doi.org/10 1997
[6]

Adam: A Method for Stochastic Optimization

Diederik P. Kingma and Jimmy Ba. “Adam: A Method for Stochastic Optimization”. In:arXiv preprint arXiv:1412.6980(2014). arXiv:1412.6980

arXiv 2014
[7]

arXiv preprint arXiv:2407.15532

Kamesh Korangi, Christophe Mues, and Cristi ´an Bravo.Large-scale time-varying portfolio opti- misation using graph attention networks. arXiv preprint arXiv:2407.15532. 2024.url:https : //arxiv.org/abs/2407.15532

arXiv 2024
[8]

Leveraging BiLSTM-GAT for enhanced stock market prediction: a dual-graph approach to portfolio optimization

Xiaobin Lu, Josiah Poon, and Matloob Khushi. “Leveraging BiLSTM-GAT for enhanced stock market prediction: a dual-graph approach to portfolio optimization”. In:Applied Intelligence55 (2025), p. 601. doi:10.1007/s10489-025-06462-w.url:https://doi.org/10.1007/s10489-025-06462- w

work page doi:10.1007/s10489-025-06462-w.url:https://doi.org/10.1007/s10489-025-06462- 2025

Show all 14 references

[9]

Accessed: 2025-07-28

MarketAux.MarketAux API.https://www.marketaux.com/. Accessed: 2025-07-28. 2025

2025
[10]

SSRN preprint

Gr ´egoire Pacreau, Edmond Lezmi, and Jiali Xu.Graph neural networks for asset management. SSRN preprint. 2021.url:https://ssrn.com/abstract=3976168

2021
[11]

Effects of Daily News Sentiment on Stock Price Forecasting

S. Srinivas et al. “Effects of Daily News Sentiment on Stock Price Forecasting”. In:arXiv preprint arXiv:2308.08549(2023).url:https://arxiv.org/abs/2308.08549

2023 arXiv
[12]

Petar Veli ˇckovi´c et al.Graph Attention Networks. 2018. arXiv:1710 . 10903 [stat.ML].url: https://arxiv.org/abs/1710.10903

2018 arXiv
[13]

arXiv preprint arXiv:2005.13665

Zihao Zhang, Stefan Zohren, and Stephen Roberts.Deep learning for portfolio optimization. arXiv preprint arXiv:2005.13665. 2020.url:https://arxiv.org/abs/2005.13665

2005 arXiv
[14]

Two-Stage Portfolio Optimization Integrating Optimal Sharpe Ratio Measure and Ensemble Learning

Zhongbao Zhou et al. “Two-Stage Portfolio Optimization Integrating Optimal Sharpe Ratio Measure and Ensemble Learning”. In:IEEE Access20 (2022).doi:10.1109/ACCESS.2022.3232281.url: https://doi.org/10.1109/ACCESS.2022.3232281. 14 7 Appendix 7.1 Feature Description Table 4: Pric...

2022

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.

[1] [1]

Accessed: 2025-07-28

Alpha Vantage Inc.Alpha Vantage API.https://www.alphavantage.co/. Accessed: 2025-07-28. 2025

2025

[2] [2]

High-performance stock index trading: making effective use of a deep long short-term memory neural network

Chariton Chalvatzis and Dimitrios Hristu-Varsakelis. “High-performance stock index trading: making effective use of a deep long short-term memory neural network”. In:arXivabs/1902.03125 (May 2019).url:https://arxiv.org/abs/1902.03125

arXiv 1902

[3] [3]

Graph neural networks for deep portfolio optimization

¨Omer Ekmekc ¸io˘glu and Mustafa C ¸ . Pınar. “Graph neural networks for deep portfolio optimization”. In:Neural Computing and Applications35 (2023), pp. 20663–20674.doi:10.1007/s00521-023- 08862-w.url:https://doi.org/10.1007/s00521-023-08862-w

work page doi:10.1007/s00521-023- 2023

[4] [4]

Louis.3-Month Treasury Bill: Secondary Market Rate (TB3MS).https: //fred.stlouisfed.org/series/TB3MS

Federal Reserve Bank of St. Louis.3-Month Treasury Bill: Secondary Market Rate (TB3MS).https: //fred.stlouisfed.org/series/TB3MS. Accessed: 2025-07-28. 2025

2025

[5] [5]

Long Short-Term Memory

Sepp Hochreiter and J¨ urgen Schmidhuber. “Long Short-Term Memory”. In:Neural Computation9.8 (Nov. 1997), pp. 1735–1780.doi:10.1162/neco.1997.9.8.1735.url:https://doi.org/10. 1162/neco.1997.9.8.1735

work page doi:10.1162/neco.1997.9.8.1735.url:https://doi.org/10 1997

[6] [6]

Adam: A Method for Stochastic Optimization

Diederik P. Kingma and Jimmy Ba. “Adam: A Method for Stochastic Optimization”. In:arXiv preprint arXiv:1412.6980(2014). arXiv:1412.6980

arXiv 2014

[7] [7]

arXiv preprint arXiv:2407.15532

Kamesh Korangi, Christophe Mues, and Cristi ´an Bravo.Large-scale time-varying portfolio opti- misation using graph attention networks. arXiv preprint arXiv:2407.15532. 2024.url:https : //arxiv.org/abs/2407.15532

arXiv 2024

[8] [8]

Leveraging BiLSTM-GAT for enhanced stock market prediction: a dual-graph approach to portfolio optimization

Xiaobin Lu, Josiah Poon, and Matloob Khushi. “Leveraging BiLSTM-GAT for enhanced stock market prediction: a dual-graph approach to portfolio optimization”. In:Applied Intelligence55 (2025), p. 601. doi:10.1007/s10489-025-06462-w.url:https://doi.org/10.1007/s10489-025-06462- w

work page doi:10.1007/s10489-025-06462-w.url:https://doi.org/10.1007/s10489-025-06462- 2025