Pith. sign in

REVIEW 3 major objections 5 minor 35 references

STN-TGAT: Top-K Portfolio Construction via Prior-Guided Graph Attention with Learnable Soft-Threshold Sparsification

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read STN-TGAT beats stock-picking baselines by training on top-ranked picks with an adaptive NMI-prior graph.

desk verdict A workmanlike incremental stock-ranking paper whose headline profitability claim is currently undermined by an unexplained discrepancy between Table 1 and Table 2 for the same model configuration. read the letter →

arxiv 2607.19385 v1 pith:VWSPAIA7 submitted 2026-07-01 cs.LG

Haoran Guo , Yutong Lu , Li Zhang This is my paper · ORCID
classification cs.LG
keywords stockrankingportfolioconstructiongraphattentionnetworktransformernormalizedmutualinformationsoft-thresholdsparsificationtop-kselectionListNet
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 argues that for realistic Top-K portfolio construction, training a model to rank stocks correctly near the top of the list matters more than predicting returns accurately, and that the graph connecting stocks should be learned from a nonlinear dependence prior with adaptive sparsification rather than fixed or dense. It proposes STN-TGAT, which combines a temporal Transformer, a graph attention network guided by a Normalized Mutual Information prior, and a learnable soft-threshold gate that suppresses weak edges. In net-of-fee backtests on the 50 largest S&P 500 constituents, the model selects a Top-5 portfolio that beats all compared baselines on IRR and Sharpe ratio. The practical stakes: a decision-aligned ranking loss plus a sparsified relational prior can turn noisy market data into investable stock picks.

What carries the argument

The load-bearing mechanism is the prior-guided, sparsified graph: an adjacency matrix $A_0$ from Normalized Mutual Information between stock return series is gated by $g_{ij} = \sigma(\beta((A_0)_{ij} - t_g))$ with learnable threshold $t_g$, yielding $A_\phi = A_0 \odot G$; this sparsified adjacency is injected into GAT attention logits as $\lambda \log((A_\phi)_{ij} + \varepsilon)$, so the data-driven attention is biased toward edges the NMI prior supports. Training uses a top-weighted ListNet loss that down-weights lower-ranked positions geometrically, an auxiliary MSE term that anchors score magnitudes, and a graph sparsity regularizer. The gating mechanism is what makes the graph adaptive: it smoothly attenuates weak dependencies rather than hard-thresholding them, and it is learned jointly with the prediction objective.

What would settle it

Run the identical Top-5, net-of-fee backtest on a longer out-of-sample window after the paper's data cutoff (e.g., all of 2025), and compare the Sharpe ratio of STN-TGAT against the GRU baseline; if the gap shrinks below one standard deviation or reverses, the central claim of consistent outperformance fails. A simpler check: report the same backtest with an S&P 500 buy-and-hold portfolio to see whether the absolute returns represent market drift.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central claim is that combining a head-weighted ListNet ranking objective with a graph attention network whose adjacency is an NMI-based prior refined by a learnable soft-threshold gate yields the best net-of-fee Top-5 portfolio among all tested models: 18.07% IRR and a 2.99 Sharpe ratio, versus 13.41% and 2.66 for the strongest recurrent baseline (GRU). The ablations support the mechanism: removing the NMI prior drops IRR from 18.07% to 8.63%, replacing NMI with GLASSO drops it to 3.81%, and disabling sparsification lowers Sharpe from 2.99 to 2.59. The paper concludes that decision-aligned training with adaptive relational modeling is a coherent and practical framework for data-driven portfolio construction.

Load-bearing premise

The claimed outperformance rests on the assumption that the short 75-day test window is representative and that the observed differences in IRR and Sharpe ratio are economically meaningful rather than statistical noise, despite overlapping standard deviations and no buy-and-hold market baseline.

Editorial extensions

If this is right

  • A top-weighted listwise loss is the main driver of portfolio performance: swapping it for pointwise or pairwise losses cuts IRR roughly by a factor of two to three in the paper's ablations, so ranking objectives should be chosen with the investment decision in mind.
  • Learnable soft-threshold sparsification improves risk-adjusted returns over a fixed NMI graph (Sharpe 2.99 vs 2.59), so adaptive graph density is a transferable component for noisy relational data.
  • The nonlinear NMI prior outperforms a linear GLASSO prior by a wide margin (Sharpe 2.99 vs 0.75), suggesting that linear dependence measures may miss relations relevant to Top-K selection.
  • Score-based weighting within the Top-5 set beats equal weighting (IRR 0.181 vs 0.152), implying predicted score magnitudes carry investable information beyond the ordering.

Reading between the lines

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

  • If the ranking-loss results generalize, the design principle—optimize for the head of the ranking, not for full-list agreement—should transfer to other decision-centric learning tasks such as recommendation, search, or candidate screening, regardless of the encoder architecture.
  • The reported Sharpe gap between STN-TGAT and GRU (2.99 vs 2.66) sits within one standard deviation, and the test window covers only about 75 trading days; a longer or repeated out-of-sample evaluation could decide whether the gap is signal or noise.
  • The absence of a buy-and-hold S&P 500 benchmark leaves the absolute 18.07% IRR unanchored; adding such a baseline would clarify whether the model earns an excess return or mostly captures market drift in the test period.
  • The soft-threshold gating of an NMI prior is a general recipe for turning dense correlation matrices into adaptive sparse graphs, so it could be applied to other financial settings (e.g., sector rotation, credit networks) or other domains with noisy relational measurements.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes STN-TGAT, a hybrid Transformer/GAT model for Top-K portfolio construction. A temporal Transformer encodes daily features, an NMI-based prior graph with learnable soft-threshold sparsification guides a GAT module, and a head-weighted ListNet loss combined with MSE and graph regularization is used for training. The model is evaluated on 50 S&P 500 constituents over 752 trading days (2022-2024) with a Top-5 long-only net-of-fee backtest. The paper claims the highest IRR (18.07%) and Sharpe ratio (2.99) among baselines, while acknowledging that LSTM achieves higher values on some pure ranking metrics. Ablations study the loss components, the NMI prior, sparsification, and within-portfolio weighting.

Significance. If the reported gains are reproducible, the paper makes a useful contribution to decision-aware ranking for portfolio construction: the explicit Top-K formulation, differentiable ranking surrogate, adaptive graph sparsification, and net-of-fee evaluation are appropriate and practical. The paper's strengths include a clean problem formulation, multiple ablations isolating loss, prior, sparsification, and allocation choices, and an explicit treatment of transaction costs. The stress-test concern about circularity is correctly rejected: the NMI prior is computed from training returns and the threshold is learned, so no self-referential construction is present. However, the empirical foundation is weakened by an internal inconsistency between Table 1 and Table 2 for the same configuration, an unsupported 'consistently outperforms' claim in the abstract, and the absence of significance testing and an index benchmark.

major comments (3)
  1. [§6.1 Table 1 vs §6.2 Table 2] The identical configuration 'ListNet+MSE' is reported with different results in Table 1 and Table 2: IRR 0.1807±0.1176 / Sharpe 2.9940±1.3287 in Table 1 versus IRR 0.2132±0.1309 / Sharpe 3.3130±1.5240 in Table 2. Since ListNet+MSE is identified as the final training objective, these two rows should be identical. The discrepancy of 3.25 percentage points in IRR and 0.319 in Sharpe directly affects the headline 'highest IRR and Sharpe' claim. Please reconcile the two tables, state which set of numbers is canonical, and explain the source of the difference (e.g., different seeds, different test split, or a typo).
  2. [§1 Abstract and §6.1 Table 1] The abstract states that STN-TGAT 'consistently outperforms benchmark models from predictive accuracy and investment profitability,' but Table 1 shows LSTM achieving higher MRR (0.2222 vs 0.1879) and RBO (0.0601 vs 0.0509) than STN-TGAT. The text in §6.1 acknowledges LSTM is higher on 'certain pure ranking metrics,' which contradicts the abstract's consistency claim. Please qualify the claim to reflect that the advantage is specifically in portfolio profitability, not in pure ranking accuracy.
  3. [§5.3 Evaluation and Table 1] No significance tests are provided, and the test window is approximately 75 trading days (10% of 752 days). The Sharpe difference between STN-TGAT (2.9940±1.3287) and GRU (2.6553±0.5632) overlaps within one standard deviation, so the claimed risk-adjusted outperformance may be noise. Additionally, no S&P 500 buy-and-hold baseline is reported, leaving the absolute returns uncontextualized. Please add hypothesis tests or bootstrap confidence intervals for the return differences, and include a broad index benchmark to make the profitability claim falsifiable.
minor comments (5)
  1. [§6.1 heading] The heading 'Ranking Accuracy and Investment Performance Comparision' contains a typo: 'Comparision' should be 'Comparison'.
  2. [Table 1 caption] The caption contains the stray phrase 'check the left tables,' which appears to be an editorial note and should be removed.
  3. [Eq. (5)] The sentence 'tg∈(0,1) is a learnable threshold parameterized' is missing a word; it should read 'a learnable threshold parameter' or 'a parameterized learnable threshold.'
  4. [§6.1] The text contains the typo 'LST Mattains' instead of 'LSTM attains'.
  5. [§4.3] The same symbol Aϕ is used both before and after symmetrization and max-scaling; please clarify the notation to avoid ambiguity about which matrix is used in Eq. (7).

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: standard supervised pipeline with held-out backtest; the headline concern is reproducibility, not derivation.

full rationale

The claimed derivation chain is self-contained. The NMI prior (Eq. 3-4) is computed from training-window historical returns and then used as an input to the GAT attention logits (Eq. 7); the soft-threshold gate (Eq. 5) and its learnable threshold are optimized through the composite objective (Eq. 15). The ranking target y_rank and regression target y_i are realized next-period returns, held out from the feature window, so the predicted scores s_i are not defined as, nor fitted to, the backtest outcome. The reported IRR and Sharpe are measured on a held-out test window rather than being implied by the loss function, and the comparison baselines share the same NMI prior for fairness. Self-citations in the related work (e.g., [30], [32]-[34]) are contextual and not load-bearing; no uniqueness theorem or prior claim by the same authors is invoked to force the architecture choice. Separately, but not a circularity issue, Tables 1 and 2 report inconsistent IRR/Sharpe for the same 'ListNet+MSE' configuration (0.1807/2.9940 vs 0.2132/3.3130), which is a reproducibility/correctness concern, not a self-referential derivation.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the validity of the NMI prior for out-of-sample dependencies, the realism of the backtest setup, and the representativeness of the short test window. No new physical entities are introduced.

free parameters (6)
  • tg (learnable soft-threshold) = learned during training
    Threshold in Eq. (5) that sparsifies the NMI adjacency; it is the key adaptive component.
  • gamma
    Decay factor for head-weighted ListNet in Eq. (11); no value reported.
  • tau_rank
    Teacher temperature in Eq. (10); no value reported.
  • beta
    Sharpness of the gating sigmoid in Eq. (5).
  • lambda
    Weight of the NMI prior term in attention logits, Eq. (7).
  • rho
    Target gating density in Eq. (14).
assumptions (4)
  • domain assumption The NMI-based dependency graph computed from historical returns remains informative for future stock relationships.
    Invoked in Section 4.3 when constructing the prior graph and in ablation analysis.
  • domain assumption The daily close-to-close rebalancing with proportional transaction costs is a realistic proxy for actual trading.
    Used in the backtest in Section 5.3; parameters deferred to the appendix.
  • domain assumption The last 10% of the sample (about 75 trading days) is representative of the model's out-of-sample performance.
    Chronological 8:1:1 split in Section 5.1; small test window underlies all significance claims.
  • domain assumption Standard ML assumptions: the training/validation/test split avoids information leakage and the holdout is unbiased.
    Section 5.1 describes winsorizing and standardization on training stats; no leakage is apparent but code is not available to verify.

how reviews work

0 comments
Cite this review

Pith. "Pith review of STN-TGAT: Top-K Portfolio Construction via Prior-Guided Graph Attention with Learnable Soft-Threshold Sparsification." pith.science (2026). https://pith.science/paper/VWSPAIA7

@misc{pith2026260719385,
  author       = {Pith},
  title        = {Pith review of: STN-TGAT: Top-K Portfolio Construction via Prior-Guided Graph Attention with Learnable Soft-Threshold Sparsification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VWSPAIA7}},
  note         = {Machine review of arXiv:2607.19385}
}
abstract

This paper tackles the problem of stock ranking and portfolio construction under realistic investment settings by jointly modeling temporal dynamics and cross-sectional dependencies. We propose the Soft-Threshold NMI-prior Transformer Graph Attention Network (STN-TGAT), which integrates a temporal Transformer with a Graph Attention Network to capture long-horizon sequential patterns and dynamic inter-stock relationships. An NMI-based prior graph combined with a soft-threshold sparsification mechanism enhances structural robustness by mitigating noisy correlations while preserving informative connections. The portfolio formation process incorporates practical considerations, including Top-5 selection within the Top-50 $S\&P$ 500 constituents, explicit weight allocation, and transaction cost adjustment, thereby aligning the evaluation with real-world trading conditions. Empirical results on real-world data demonstrate that STN-TGAT consistently outperforms benchmark models from predictive accuracy and investment profitability measured by portfolio returns. These findings suggest that combining decision-aligned training with adaptive relational modeling provides a coherent and practically effective framework for data-driven portfolio construction.

Figures

Figures reproduced from arXiv: 2607.19385 by the authors.

Figure 1
Figure 1. Performance comparison of different methods [PITH_FULL_IMAGE:figures/full_fig_p011_1.png] view at source ↗
Figure 2
Figure 2. Ablation study visualization: (a) and (b) show the distributions of IRR [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. Analysis of the learned adjacency. Panel (a) reports the distribution of learned edge weights, (b) shows the cumulative distribution, and (c) compares prior weights A0 with learned weights Aϕ; the y=x diagonal indicates perfect agreement. (a) Delta heatmap: Aϕ − A0 (b) Learned Aϕ heatmap (reordered) [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Heatmap diagnostics of the learned graph vs. the NMI prior. Panel (a) visualizes adjustments to the prior via the difference Aϕ−A0, and (b) shows the learned adjacency after node reordering by aggregate edge weight. The structural diagnostics are consistent with this o…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

35 extracted references · 32 canonical work pages

  1. [1]

    Forecasting financial market structure from net- work features using machine learning.Knowledge and Information Systems, 2024

    Douglas Castilho, Tharsis TP Souza, Soong Moon Kang, João Gama, and André CPLF de Carvalho. Forecasting financial market structure from net- work features using machine learning.Knowledge and Information Systems, 2024

    work page 2024

  2. [2]

    Dynamic deep convolutional candle- stick learner.arXiv preprint arXiv:2201.08669, 2022

    Jun-Hao Chen and Yun-Cheng Tsai. Dynamic deep convolutional candle- stick learner.arXiv preprint arXiv:2201.08669, 2022

    arXiv 2022

  3. [3]

    A novel convolutional neural networks for stock trading based on ddqn algorithm

    Kai Cui, Ruizhe Hao, Yuling Huang, Jianqing Li, and Yunlin Song. A novel convolutional neural networks for stock trading based on ddqn algorithm. IEEE Access, 2023

    work page 2023

  4. [4]

    Hierar- chical multi-scale gaussian transformer for stock movement prediction

    Qianggang Ding, Sifan Wu, Hao Sun, Jiadong Guo, and Jian Guo. Hierar- chical multi-scale gaussian transformer for stock movement prediction. In IJCAI, 2020

    work page 2020

  5. [5]

    Relation-aware dynamic attributed graph attention network for stocks rec- ommendation.Pattern Recognition, 2022

    Shibo Feng, Chen Xu, Yu Zuo, Guo Chen, Fan Lin, and Jianbing XiaHou. Relation-aware dynamic attributed graph attention network for stocks rec- ommendation.Pattern Recognition, 2022

    work page 2022

  6. [6]

    Graph-based stock recommendation by time-aware relational attention network.ACM Transactions on Knowledge Discovery from Data (TKDD), 2021

    Jianliang Gao, Xiaoting Ying, Cong Xu, Jianxin Wang, Shichao Zhang, and Zhao Li. Graph-based stock recommendation by time-aware relational attention network.ACM Transactions on Knowledge Discovery from Data (TKDD), 2021

    work page 2021

  7. [7]

    Dystage: Dynamic graph representation learning for asset pricing via spatio-temporal atten- tion and graph encodings

    Jingyi Gu, Junyi Ye, Ajim Uddin, and Guiling Wang. Dystage: Dynamic graph representation learning for asset pricing via spatio-temporal atten- tion and graph encodings. InProceedings of the 5th ACM International Conference on AI in Finance, 2024

    work page 2024

  8. [8]

    Static-dynamic graph neural network for stock recommendation

    Yanshen He, Qiutong Li, Feng Wu, and Jianliang Gao. Static-dynamic graph neural network for stock recommendation. InProceedings of the 34th international conference on scientific and statistical database management, 2022

    work page 2022

Show all 35 references
  1. [9]

    A study of dynamic stock relationship modeling and s&p500 price forecasting based on differential graph transformer.arXiv preprint arXiv:2506.18717, 2025

    Linyue Hu and Qi Wang. A study of dynamic stock relationship modeling and s&p500 price forecasting based on differential graph transformer.arXiv preprint arXiv:2506.18717, 2025

    2025 arXiv

  2. [10]

    Stock price prediction methodology using random forest algorithm and sup- port vector machine.Materials Today: Proceedings, 2022

    PavanKumarIlla,BalakesavareddyParvathala,andAnandKumarSharma. Stock price prediction methodology using random forest algorithm and sup- port vector machine.Materials Today: Proceedings, 2022

    2022

  3. [11]

    Gcnet: Graph-based prediction of stock price movement using graph convolutional network.Engineering Ap- plications of Artificial Intelligence, 2022

    Alireza Jafari and Saman Haratizadeh. Gcnet: Graph-based prediction of stock price movement using graph convolutional network.Engineering Ap- plications of Artificial Intelligence, 2022

    2022

  4. [12]

    Stock price prediction using a frequency decomposition based gru transformer neural network.Applied Sciences, 2022

    Chengyu Li and Guoqi Qian. Stock price prediction using a frequency decomposition based gru transformer neural network.Applied Sciences, 2022

    2022

  5. [13]

    Vgc-gan: A multi- graph convolution adversarial network for stock price prediction.Expert Systems with Applications, 2024

    Dongbo Ma, Da Yuan, Maojun Huang, and Ling Dong. Vgc-gan: A multi- graph convolution adversarial network for stock price prediction.Expert Systems with Applications, 2024. STN-TGAT 17

    2024

  6. [14]

    Stock market forecasting using the random forest and deepneuralnetworkmodelsbeforeandduringthecovid-19period.Frontiers in Environmental Science, 2022

    Abdullah Bin Omar, Shuai Huang, Anas A Salameh, Haris Khurram, and Muhammad Fareed. Stock market forecasting using the random forest and deepneuralnetworkmodelsbeforeandduringthecovid-19period.Frontiers in Environmental Science, 2022

    2022

  7. [15]

    Forecast- ing stock indices: Stochastic and artificial neural network models.Compu- tational Economics, 2025

    Naman Krishna Pande, Arun Kumar, and Arvind Kumar Gupta. Forecast- ing stock indices: Stochastic and artificial neural network models.Compu- tational Economics, 2025

    2025

  8. [16]

    A systematic reviewongraphneuralnetwork-basedmethodsforstockmarketforecasting

    Manali Patel, Krupa Jariwala, and Chiranjoy Chattopadhyay. A systematic reviewongraphneuralnetwork-basedmethodsforstockmarketforecasting. ACM Computing Surveys, 2024

    2024

  9. [17]

    An improved convolutional recurrent neural network for stock price forecasting.IAES International Journal of Artificial Intelligence, 2024

    HV Pham, HP Lam, LN Duy, TB Pham, and TD Trinh. An improved convolutional recurrent neural network for stock price forecasting.IAES International Journal of Artificial Intelligence, 2024

    2024

  10. [18]

    Forecasting stock market movement direction using sentiment analysis and support vector machine

    Rui Ren, Desheng Dash Wu, and Tianxiang Liu. Forecasting stock market movement direction using sentiment analysis and support vector machine. IEEE Systems Journal, 2018

    2018

  11. [19]

    Stock ranking prediction using list-wise approach and node embedding technique.IEEE Access, 2021

    Suman Saha, Junbin Gao, and Richard Gerlach. Stock ranking prediction using list-wise approach and node embedding technique.IEEE Access, 2021

    2021

  12. [20]

    Ex- ploring the scale-free nature of stock markets: Hyperbolic graph learning for algorithmic trading

    Ramit Sawhney, Shivam Agarwal, Arnav Wadhwa, and Rajiv Shah. Ex- ploring the scale-free nature of stock markets: Hyperbolic graph learning for algorithmic trading. InProceedings of the Web Conference 2021, 2021

    2021

  13. [21]

    Learn- ing dynamic dependencies with graph evolution recurrent unit for stock predictions.IEEE Transactions on Systems, Man, and Cybernetics: Sys- tems, 2023

    Hu Tian, Xingwei Zhang, Xiaolong Zheng, and Daniel Dajun Zeng. Learn- ing dynamic dependencies with graph evolution recurrent unit for stock predictions.IEEE Transactions on Systems, Man, and Cybernetics: Sys- tems, 2023

    2023

  14. [22]

    Graph denoising networks: a deep learning framework for equity portfolio construction

    Edward Turner and Mihai Cucuringu. Graph denoising networks: a deep learning framework for equity portfolio construction. InProceedings of the Fourth ACM International Conference on AI in Finance, 2023

    2023

  15. [23]

    Distance correlation market graph: The case of s&p500 stocks.Mathematics, 2023

    Samuel Ugwu, Pierre Miasnikof, and Yuri Lawryshyn. Distance correlation market graph: The case of s&p500 stocks.Mathematics, 2023

    2023

  16. [24]

    Attention is all you need.Advances in neural information processing systems, 2017

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need.Advances in neural information processing systems, 2017

    2017

  17. [25]

    Graphattentionnetworks

    Petar Veličković, Guillem Cucurull, Arantxa Casanova, Adriana Romero, PietroLio,YoshuaBengio,etal. Graphattentionnetworks. InInternational conference on learning representations, 2018

    2018

  18. [26]

    Mg- conv: A spatiotemporal multi-graph convolutional neural network for stock market index trend prediction.Computers and Electrical Engineering, 2022

    Changhai Wang, Hui Liang, Bo Wang, Xiaoxu Cui, and Yuwei Xu. Mg- conv: A spatiotemporal multi-graph convolutional neural network for stock market index trend prediction.Computers and Electrical Engineering, 2022

    2022

  19. [27]

    Stock market index prediction using deep transformer model.Expert Systems with Applications, 2022

    Chaojie Wang, Yuanyuan Chen, Shuqi Zhang, and Qiuhui Zhang. Stock market index prediction using deep transformer model.Expert Systems with Applications, 2022

    2022

  20. [28]

    Hist: A graph-based framework for stock trend fore- casting via mining concept-oriented shared information.arXiv preprint arXiv:2110.13716, 2021

    Wentao Xu, Weiqing Liu, Lewen Wang, Yingce Xia, Jiang Bian, Jian Yin, and Tie-Yan Liu. Hist: A graph-based framework for stock trend fore- casting via mining concept-oriented shared information.arXiv preprint arXiv:2110.13716, 2021. 18 H. Guo et al

    arXiv

  21. [29]

    Forecasting stock prices using stock correlation graph: A graph convolu- tional network approach

    Xingkun Yin, Da Yan, Abdullateef Almudaifer, Sibo Yan, and Yang Zhou. Forecasting stock prices using stock correlation graph: A graph convolu- tional network approach. In2021 International Joint Conference on Neural Networks (IJCNN), 2021

    2021

  22. [30]

    DGDNN:Decoupledgraphdiffusionneuralnetworkforstockmovementpre- diction

    Zinuo You, Zijian Shi, Hongbo Bo, John Cartlidge, Li Zhang, and Yan Ge. DGDNN:Decoupledgraphdiffusionneuralnetworkforstockmovementpre- diction. InProceedings of the 16th International Conference on Agents and Artificial Intelligence (ICAART 2024) – Volume 2, pages 431–442, 2024

    2024

  23. [31]

    Stock price prediction and comparative analysis using rnn, lstm and arima

    Qianhui Yu. Stock price prediction and comparative analysis using rnn, lstm and arima. InAIP Conference Proceedings, 2024

    2024

  24. [32]

    A feature-importance-aware and robust aggre- gator for gcn

    Li Zhang and Haiping Lu. A feature-importance-aware and robust aggre- gator for gcn. InProceedings of the 29th ACM International Conference on Information & Knowledge Management, pages 1813–1822, 2020

    2020

  25. [33]

    Hop-hop relation-aware graph neural networks.arXiv preprint arXiv:2012.11147, 2020

    Li Zhang, Yan Ge, and Haiping Lu. Hop-hop relation-aware graph neural networks.arXiv preprint arXiv:2012.11147, 2020

    2012 arXiv

  26. [34]

    Node-feature convolution for graph convolutional networks.Pattern Recognition, 128: 108661, 2022

    Li Zhang, Heda Song, Nikolaos Aletras, and Haiping Lu. Node-feature convolution for graph convolutional networks.Pattern Recognition, 128: 108661, 2022

    2022

  27. [35]

    Relational temporal graph convolutional networks for ranking-based stock prediction

    Zetao Zheng, Jie Shao, Jia Zhu, and Heng Tao Shen. Relational temporal graph convolutional networks for ranking-based stock prediction. In2023 IEEE 39th International Conference on Data Engineering (ICDE), 2023

    2023

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.