Pith. sign in

REVIEW 5 major objections 5 minor 5 cited by

HATS: A Hierarchical Graph Attention Network for Stock Movement Prediction

T0 review · 5 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a hierarchical graph attention network which learns to weight corporate relations can improve stock movement prediction for both individual stocks and market indices, beating existing baselines on Sharpe ratio and…

desk verdict The empirical headline doesn't survive contact with the tables, but the core idea—learned per-relation weighting for stock prediction—is worth a careful look. read the letter →

arxiv 1908.07999 v3 pith:G3SVQX4Z submitted 2019-08-07 q-fin.ST cs.AIcs.CE

Raehyun Kim , Chan Ho So , Minbyul Jeong , Sanghoon Lee , Jinkyu Kim , Jaewoo Kang This is my paper · ORCID
classification q-fin.STcs.AIcs.CE
keywords stockmovementpredictiongraphneuralnetworkattentionmechanismcorporaterelationsmarketindexSharperatioWikidatameta-pathhierarchical
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 when predicting stock movements from a corporate relational graph, different types of company relations carry very different predictive value, and indiscriminately aggregating them hurts performance. It proposes HATS, a hierarchical attention network that first weights neighbors within each relation type and then weights relation types themselves, so the model learns which relations matter for the prediction task. The authors show that this selective aggregation outperforms existing graph-based and non-graph baselines on both individual stock classification and market index classification, with 19.8% higher Sharpe ratio and 3% higher F1 score on average. The central message is that relation selection should be learned, not fixed, because the usefulness of a relation depends on the market phase.

What carries the argument

The central mechanism is the two-layer hierarchical attention: a state attention layer computes weights over neighboring companies within each relation type, producing a summarized representation per relation, and a relation attention layer then weights these relation summaries based on the target company's current state, with the weighted summaries added to the node's own representation. The state- and relation-dependent weights allow the model to filter noisy relations and focus on informative ones, in contrast to static adjacency-matrix methods like GCN or the temporal graph convolution of TGC. This machinery carries the paper's claim that automatic relation selection improves prediction.

What would settle it

Train HATS on a graph whose relation type labels are randomly permuted, or whose edges are rewired while preserving degree; if the improvement in Sharpe ratio and F1 score over the baselines persists unchanged, then the gains come from the attention mechanism's extra capacity rather than from selectively aggregating meaningful relation types, and the paper's central claim would be falsified. A second check: if the learned relation attention weights show no correspondence to the per-relation predictive performance measured in Section 5.2, the selection mechanism is not doing the work.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that stock movement prediction improves when a model is able to selectively aggregate information from different corporate relation types. It demonstrates that the choice of relational data changes prediction performance dramatically: some relations, such as parent-organization and subsidiary ties, help, while dense geographical relations such as country of origin degrade accuracy. The paper's contribution is HATS, a two-level attention module that summarizes neighbor information per relation and then weights the relation summaries, adding the result to each company's representation. On S&P 500 data, HATS outperforms existing methods, including a manually-selected-top-20-relations GCN, on F1 score, and achieves the highest average Sharpe ratio among all tested models. The same module is adapted to predict market index movements through graph pooling, outperforming baselines there as well.

Load-bearing premise

The entire result rests on the assumption that the Wikidata-derived corporate graph, using meta-paths of at most two hops and excluding companies with no relations, captures the relationships that actually drive stock prices.

Editorial extensions

If this is right

  • Using all available relations without selection can degrade prediction; models that learn to weight relations adaptively achieve the best results.
  • The same relational modeling module transfers from node classification (individual stock movement) to graph classification (market index movement) when combined with graph pooling.
  • HATS removes the need for manual selection of relation types, since its attention weights learn which relations matter per market phase.
  • The profitability gains, measured by average daily return and Sharpe ratio, are stronger for HATS than for any baseline, indicating that attention-based relation selection improves risk-adjusted returns.

Reading between the lines

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

  • The finding that dense, semantically weak relations hurt performance suggests a testable preprocessing rule: filtering or down-weighting high-degree relation types could improve even non-attention baselines like GCN.
  • If HATS's gains persist on graphs built from richer or alternative knowledge bases, or from text-extracted relations, the method would become a general-purpose relational modeling module for financial graphs.
  • The learned relation attention weights could be used as interpretable indicators of which corporate connections matter at a given time, potentially serving as features in fundamental or risk analysis.
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

5 major / 5 minor

Summary. The paper proposes HATS, a hierarchical graph attention network for stock movement prediction. Node representations are initialized from an LSTM/GRU feature extractor over price change rates; HATS then applies two levels of attention—state-level attention over neighbors within each relation type and relation-level attention over relation types—to produce updated node representations, which are used for individual stock classification (node classification) and market index movement prediction via mean pooling (graph classification). The authors construct a corporate graph from Wikidata, using direct relations and two-hop meta-paths among 431 S&P 500 companies, and evaluate over 12 phases (text says 8) comparing against MLP, CNN, LSTM, GCN, GCN-TOP20, and TGC. They report that HATS outperforms all baselines in F1, accuracy, average daily return, and Sharpe ratio, and they analyze which relation types are useful.

Significance. The contribution is potentially useful: it is among the first to explicitly model relation-type selection for stock prediction and to cast index prediction as graph classification, with code and data publicly released. The relation-level analysis (Table 1 and Figure 5) is a valuable empirical observation that not all corporate relations help prediction. However, the central claim of consistent superiority is not currently supported by the evidence as presented: the experimental reporting contains internal inconsistencies, an unverified headline performance number, a baseline constructed with test-set leakage, and no uncertainty quantification. With corrected analysis the work could be a solid contribution, but the quantitative claims in the current manuscript are not yet established.

major comments (5)
  1. [Section 5.1, Table 2] In Table 2, the F1 rows for Phase 1 and Phase 7 are identical to four decimal places for all seven models, as are the F1 rows for Phase 2 and Phase 8, even though the corresponding accuracy rows differ. Exact equality across all models for two different market windows is implausible and indicates a copying or data-processing error. Since Table 2 is the primary evidence for the F1 improvement claimed in the abstract, the table must be corrected and the source of the duplication explained.
  2. [Section 5.1, Figure 3, Tables 2-3] The text states 'we divided our entire dataset into 8 smaller datasets' and that 'each phase consists of 250 days of training, 50 days of evaluation, and 100 days of testing,' yet Tables 2 and 3 report 12 phases. With 1,174 trading days, eight non-overlapping 400-day phases are impossible; the degree of overlap or alternative splitting scheme is not described. The experimental timeline is therefore not reproducible as written, and the text/table mismatch must be resolved.
  3. [Section 5.2, Table 1] GCN-TOP20 is constructed by selecting the 20 best-performing relation types based on F1 scores on the Phase 4 test set (Table 1 is described as 'test set of Phase 4'). Using test-set labels to select the baseline's relation set leaks Phase 4 test information into that baseline's configuration, and also means GCN-TOP20 is not an 'existing method' as claimed. Relation selection should be performed on training/validation data only, or GCN-TOP20 should be re-framed as an oracle-style upper bound; as presented, the comparison is biased in GCN-TOP20's favor and cannot be used to support the claim that HATS outperforms all existing methods.
  4. [Section 5.1, Tables 2-3] The paper states all experiments were repeated five times and averaged, but no standard deviations, confidence intervals, or significance tests are reported for any metric. The average F1 advantage of HATS over GCN-TOP20 is 0.0069 (approximately 2.1%), and per-phase Sharpe ratios for HATS range from 0.41 to 4.39, so without variance information the reported margins may be within noise. Report per-run results with error bars and test whether HATS's advantages over LSTM, TGC, and GCN-TOP20 are statistically significant.
  5. [Abstract, Section 1, Table 3] The abstract and contribution list state that HATS's Sharpe ratio was 19.8% higher than existing baselines. This number does not match any comparison in Table 3: HATS's average Sharpe (1.9914) is about 73% higher than LSTM (1.1523), 99% higher than TGC (1.0026), and 108% higher than GCN-TOP20 (0.9589). The 3% F1 figure also does not exactly correspond to any shown comparison (HATS vs. GCN-TOP20 is 2.1%; vs. LSTM is 5.8%). The baseline and computation behind the headline numbers must be specified, or the numbers corrected.
minor comments (5)
  1. [Section 3.4] Equation numbering is duplicated: Eq. (3.9) is used for both the node-classification loss in Section 3.3 and the mean-pooling formula in Section 3.4; renumber the equations.
  2. [Section 5.4] The text refers to 'table 5' for the market index prediction results, but those results appear in Table 4; the appendix Table 5 contains relation definitions, so the cross-reference should be corrected.
  3. [Section 4.2] There is a duplicated word in 'which we we leave for future work'; this should read 'which we leave for future work.'
  4. [Table 1] Table 1 lists 'Country-Country of origin' twice with different F1 scores (0.2948 and 0.2851); one entry is presumably 'Country of origin-Country' and should be corrected.
  5. [Section 5.4] The sentence beginning 'With only five assets' starts with a lowercase 'With' after a period; this should be capitalized.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: HATS is defined by explicit equations and evaluated against external baselines on held-out phases; the main concerns are statistical reporting, not derivation circularity.

full rationale

The claimed derivation chain is self-contained. HATS is defined by Eqs. (3.1)-(3.7) as a two-level attention mechanism over relation-specific neighbor sets, using node representations from an LSTM/GRU feature extraction module; the prediction layers are linear transforms (Eq. 3.8) and graph pooling is simple mean pooling (Eqs. 3.9-3.11). No equation defines the model output in terms of the target F1 or Sharpe metric, nor is any fitted parameter renamed as a prediction. The experimental comparison is against external baselines (MLP, CNN, LSTM, GCN, TGC) on held-out test phases, so the central 'outperforms baselines' claim is not forced by construction. The only self-citation, [19] (Self-Attention Graph Pooling, same research group), is a passing reference to pooling methods and is not used in the experiments, which explicitly use mean pooling; it is therefore not load-bearing. The GCN-TOP20 baseline is built by selecting the 20 best relation types from Phase 4 test-set F1 scores (Section 5.2), which is a methodological weakness (test-set relation selection for a baseline) but not circularity in HATS's derivation. The reported 19.8% Sharpe improvement does not match any comparison in Table 3, and Table 2 contains exact duplicate rows across phases; these are correctness and statistical reporting issues, not circular derivation. Accordingly, no circular step can be exhibited from the paper's own equations or citations, and the paper is assessed as having no significant circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 1 invented entities

The central claim rests on a standard supervised-learning setup: labels defined by unstated thresholds, a manually built Wikidata graph, a fixed evaluation schedule, and a no-transaction-cost trading simulation. There are no invented physical entities. The free parameters are mostly architecture and evaluation choices that are not varied or reported. The most consequential input is the graph itself, whose construction is acknowledged to be improvable.

free parameters (6)
  • up/neutral/down label thresholds = not specified
    Section 5.1 states 'two threshold values were used' to define the three classes, but the actual values are never given; classification results depend directly on them.
  • lookback window = 50 days
    Section 5.1 sets a 50-day lookback for all models with no sensitivity analysis; results are known to depend on this.
  • portfolio size for long/short = 15 stocks
    Section 5.1: 'We selected 15 companies with the highest up class probability' for long and 15 for short; this choice affects Sharpe ratio and return comparisons.
  • GCN-TOP20 selected relations = 20 relations with highest Phase 4 test F1
    Section 5.2 selects top relations based on F1 on the Phase 4 test set; this baseline is therefore fitted to the test outcome.
  • relation embedding dimension d = not specified
    Equations (3.1)-(3.4) use a relation embedding vector of dimension d, but d is never reported; tuning this changes attention behavior.
  • train/eval/test phase lengths = 250/50/100 days
    Section 5.1 uses phases of 250 training, 50 evaluation, and 100 test days without sensitivity analysis, and the text says 8 phases while tables show 12.
assumptions (5)
  • domain assumption Stock prices are predictable to a meaningful degree from historical price changes and corporate graph structure.
    The paper cites the predictability debate [21] but designs a predictive task and profitable-portfolio evaluation that presumes predictability; Section 1.
  • domain assumption Historical price change rates are a sufficient node feature representation.
    Section 4.1 uses only Rt=(Pt-Pt-1)/Pt-1 following [14]; no fundamental or textual features are used, so the claim of relation-based improvement is contingent on this feature choice.
  • ad hoc to paper Wikidata meta-paths with at most two hops capture the economically meaningful relations among companies.
    Section 4.2 builds the homogeneous company graph from Wikidata and 75 relation/meta-path types, dropping 69 S&P 500 companies with no relations; the same section says the construction 'can be much improved'.
  • domain assumption The 250/50/100-day phase division is representative of different market conditions.
    Section 5.1 splits the 2013-2019 sample into phases to test drawdown robustness, but no test for stability of the reported improvements across alternative splits is provided.
  • domain assumption A trading strategy ignoring transaction costs is a valid profitability measure.
    The return and Sharpe computations in Equations (5.1)-(5.2) assume frictionless trading; with real costs, reported Sharpe ratios would shrink.
invented entities (1)
  • None
    purpose: No new physical or conceptual entities postulated
    All inputs are real price data and Wikidata relations; the attention weights are learned parameters, not invented entities.

how reviews work

0 comments
Cite this review

Pith. "Pith review of HATS: A Hierarchical Graph Attention Network for Stock Movement Prediction." pith.science (2026). https://pith.science/paper/G3SVQX4Z

@misc{pith2026190807999,
  author       = {Pith},
  title        = {Pith review of: HATS: A Hierarchical Graph Attention Network for Stock Movement Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/G3SVQX4Z}},
  note         = {Machine review of arXiv:1908.07999}
}
read the original abstract

Many researchers both in academia and industry have long been interested in the stock market. Numerous approaches were developed to accurately predict future trends in stock prices. Recently, there has been a growing interest in utilizing graph-structured data in computer science research communities. Methods that use relational data for stock market prediction have been recently proposed, but they are still in their infancy. First, the quality of collected information from different types of relations can vary considerably. No existing work has focused on the effect of using different types of relations on stock market prediction or finding an effective way to selectively aggregate information on different relation types. Furthermore, existing works have focused on only individual stock prediction which is similar to the node classification task. To address this, we propose a hierarchical attention network for stock prediction (HATS) which uses relational data for stock market prediction. Our HATS method selectively aggregates information on different relation types and adds the information to the representations of each company. Specifically, node representations are initialized with features extracted from a feature extraction module. HATS is used as a relational modeling module with initialized node representations. Then, node representations with the added information are fed into a task-specific layer. Our method is used for predicting not only individual stock prices but also market index movements, which is similar to the graph classification task. The experimental results show that performance can change depending on the relational data used. HATS which can automatically select information outperformed all the existing methods.

Figures

Figures reproduced from arXiv: 1908.07999 by the authors.

Figure 1
Figure 1. General framework of stock prediction using relational data. LSTM especially when a model had more layers. For this reason, we use LSTM for the individual stock prediction task and GRU for the index movement prediction task where an additional graph pooling layer is needed. Relational Modeling Module A relational modeling mod￾ule is a node updating function. Gilmer et al. considered graph-based learning as informati… view at source ↗
Figure 2
Figure 2. Hierarchical Attention Network for Stock Prediction node representations of i and j into a vector where ∈ [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (4 more)
Figure 3
Figure 3. Figure 3: Dataset arrangement for the experiment. A 50-day period is used for evaluation right before the test period. A black line indicates an actual S&P 500 Index closing price. to convert an originally heterogeneous graph into a homo￾geneous graph with only company nodes. St…
Figure 4
Figure 4. Figure 4: Comparison of different prediction models and their changes in asset value. The asset value is assumed to start at 100. the best. Therefore, we compare the results of the models with a relational modeling module and the result of LSTM. In terms of accuracy, all models …
Figure 5
Figure 5. Figure 5: Visualization of attention scores of different relations. 20 relations with highest attention scores on average and 10 relations with the lowest scores on average [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: Visualization of node representations using T-SNE. nism of HATS. As previously mentioned, HATS is designed to gather information from only useful relations. For our first case study, we calculated the attention score of each relation. By analyzing the relation types wi…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. LLM-Enhanced Dynamic Financial Knowledge Graphs for Cross-Entity Signal Propagation and alpha discovery

    stat.AP 2026-07 conditional novelty 6.0 of 10

    In controlled simulations, community-aware propagation of LLM event signals on dynamic financial knowledge graphs recovers latent communities and prices incrementally beyond direct signals, though live alpha remains untested.

  2. Pre-training Time Series Models with Stock Data Customization

    cs.CE 2025-06 conditional novelty 6.0 of 10

    Stock Specialized Pre-trained Transformer (SSPT) uses stock code classification, sector classification, and moving average prediction as pre-training tasks, and reports improved stock selection returns and Sharpe rati...

  3. From Votes to Volatility Predicting the Stock Market on Election Day

    q-fin.CP 2024-12 reject novelty 4.0 of 10

    Adding hand-coded candidate impact scores and a randomly assigned candidate context to StockMixer produces no robust improvement over the baseline on a single day of S&P 500 data.

  4. Higher Order Transformers: Enhancing Stock Movement Prediction On Multimodal Time-Series Data

    cs.LG 2024-12 conditional novelty 4.0 of 10

    A factorized 'higher-order' transformer with kernelized linear attention and tweet plus price inputs reaches 72.94% accuracy and 0.516 MCC on StockNet, behind only NL-LSTM among the baselines compared.

  5. Dynamic Graph Representation with Contrastive Learning for Financial Market Prediction: Integrating Temporal Evolution and Static Relations

    cs.LG 2024-12 conditional novelty 4.0 of 10

    A contrastive learning framework that blends dynamic stock graphs with static company relations improves next-day stock movement classification by roughly two accuracy points over four baselines.

Reference graph

Works this paper leans on

30 extracted references · 27 canonical work pages · cited by 5 Pith papers

  1. [1]

    Comparison of arimaandartificialneuralnetworksmodelsforstockpriceprediction

    Adebiyi, A.A., Adewumi, A.O., Ayo, C.K., 2014. Comparison of arimaandartificialneuralnetworksmodelsforstockpriceprediction. Journal of Applied Mathematics 2014

    work page 2014

  2. [2]

    State-of-the-art in stock prediction techniques

    Agrawal, J., Chourasia, V., Mittra, A., 2013. State-of-the-art in stock prediction techniques. International Journal of Advanced Research in Electrical, Electronics and Instrumentation Engineering 2, 1360– 1366

    work page 2013

  3. [3]

    A deep learning framework for fi- nancial time series using stacked autoencoders and long-short term memory

    Bao, W., Yue, J., Rao, Y., 2017. A deep learning framework for fi- nancial time series using stacked autoencoders and long-short term memory. PloS one 12, e0180944

    work page 2017

  4. [4]

    Twitter mood predicts the stock market

    Bollen, J., Mao, H., Zeng, X., 2011. Twitter mood predicts the stock market. Journal of computational science 2, 1–8

    work page 2011

  5. [5]

    Stock return pre- dictability and variancerisk premia: statistical inference and interna- tional evidence

    Bollerslev, T., Marrone, J., Xu, L., Zhou, H., 2014. Stock return pre- dictability and variancerisk premia: statistical inference and interna- tional evidence. Journal of Financial and Quantitative Analysis 49, 633–661

    work page 2014

  6. [6]

    FastGCN:Fastlearningwithgraph convolutional networks via importance sampling, in: International Conference on Learning Representations

    Chen,J.,Ma,T.,Xiao,C.,2018a. FastGCN:Fastlearningwithgraph convolutional networks via importance sampling, in: International Conference on Learning Representations

    work page

  7. [7]

    Chen,Y.,Wei,Z.,Huang,X.,2018b. Incorporatingcorporationrela- tionship via graph convolutional neural networks for stock price pre- diction, in: Proceedings of the 27th ACM International Conference on Information and Knowledge Management, ACM. pp. 1655–1658

    work page

  8. [8]

    Journal of Financial Economics 61, 77–106

    Dechow,P.M.,Hutton,A.P.,Meulbroek,L.,Sloan,R.G.,2001.Short- sellers, fundamental analysis, and stock returns. Journal of Financial Economics 61, 77–106

    work page 2001

Show all 30 references
  1. [9]

    Computational learning techniques for intraday fx trading using pop- ular technical indicators

    Dempster, M.A., Payne, T.W., Romahi, Y., Thompson, G.W., 2001. Computational learning techniques for intraday fx trading using pop- ular technical indicators. IEEE Transactions on neural networks 12, 744–754

    2001

  2. [10]

    Deeplearningforevent- driven stock prediction, in: Twenty-Fourth International Joint Con- ference on Artificial Intelligence

    Ding,X.,Zhang,Y.,Liu,T.,Duan,J.,2015. Deeplearningforevent- driven stock prediction, in: Twenty-Fourth International Joint Con- ference on Artificial Intelligence

    2015

  3. [11]

    Knowledge-driven event embedding for stock prediction, in: Proceedings of COLING 2016, the 26th International Conference on Computational Linguis- tics: Technical Papers, pp

    Ding, X., Zhang, Y., Liu, T., Duan, J., 2016. Knowledge-driven event embedding for stock prediction, in: Proceedings of COLING 2016, the 26th International Conference on Computational Linguis- tics: Technical Papers, pp. 2133–2142

    2016

  4. [12]

    Dong, Y., Chawla, N.V., Swami, A., 2017. metapath2vec: Scalable representation learning for heterogeneous networks, in: Proceedings of the 23rd ACM SIGKDD international conference on knowledge discovery and data mining, ACM. pp. 135–144

    2017

  5. [13]

    Tem- poral relational ranking for stock prediction

    Feng, F., He, X., Wang, X., Luo, C., Liu, Y., Chua, T.S., 2019. Tem- poral relational ranking for stock prediction. ACM Transactions on Information Systems (TOIS) 37, 27

    2019

  6. [14]

    Deep learning with long short-term memorynetworksforfinancialmarketpredictions

    Fischer, T., Krauss, C., 2018. Deep learning with long short-term memorynetworksforfinancialmarketpredictions. EuropeanJournal of Operational Research 270, 654–669

    2018

  7. [15]

    Gilmer, J., Schoenholz, S.S., Riley, P.F., Vinyals, O., Dahl, G.E.,

  8. [16]

    Inductive representation learning on large graphs, in: Advances in Neural Information Pro- cessing Systems, pp

    Hamilton, W., Ying, Z., Leskovec, J., 2017. Inductive representation learning on large graphs, in: Advances in Neural Information Pro- cessing Systems, pp. 1024–1034

    2017

  9. [17]

    Semi-supervised classification with graph convolutional networks

    Kipf, T.N., Welling, M., 2016. Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv:1609.02907

    2016 arXiv

  10. [18]

    Imagenet clas- sification with deep convolutional neural networks, in: Advances in neural information processing systems, pp

    Krizhevsky, A., Sutskever, I., Hinton, G.E., 2012. Imagenet clas- sification with deep convolutional neural networks, in: Advances in neural information processing systems, pp. 1097–1105

    2012

  11. [19]

    Self-attention graph pooling

    Lee, J., Lee, I., Kang, J., 2019. Self-attention graph pooling. arXiv preprint arXiv:1904.08082

    2019 arXiv

  12. [20]

    News impact on stock price return via sentiment analysis

    Li, X., Xie, H., Chen, L., Wang, J., Deng, X., 2014. News impact on stock price return via sentiment analysis. Knowledge-Based Systems 69, 14–23

    2014

  13. [21]

    The efficient market hypothesis and its critics

    Malkiel, B.G., 2003. The efficient market hypothesis and its critics. Journal of economic perspectives 17, 59–82

    2003

  14. [22]

    Predicting stock andstockpriceindexmovementusingtrenddeterministicdataprepa- ration and machine learning techniques

    Patel, J., Shah, S., Thakkar, P., Kotecha, K., 2015. Predicting stock andstockpriceindexmovementusingtrenddeterministicdataprepa- ration and machine learning techniques. Expert Systems with Appli- cations 42, 259–268

    2015

  15. [23]

    Stock return fore- casting: somenewevidence

    Phan, D.H.B., Sharma, S.S., Narayan, P.K., 2015. Stock return fore- casting: somenewevidence. InternationalReviewofFinancialAnal- ysis 40, 38–51

    2015

  16. [24]

    Recurrent neural net- work and a hybrid model for prediction of stock returns

    Rather, A.M., Agarwal, A., Sastry, V., 2015. Recurrent neural net- work and a hybrid model for prediction of stock returns. Expert Sys- tems with Applications 42, 3234–3241

    2015

  17. [25]

    Graph attention networks

    Veli/uni010Dkovi/uni0107, P., Cucurull, G., Casanova, A., Romero, A., Lio, P., Bengio, Y., 2017. Graph attention networks. arXiv preprint arXiv:1710.10903

    2017 arXiv

  18. [26]

    Wikidata: a free collaborative knowledge base

    Vrande/uni010Di/uni0107, D., Krötzsch, M., 2014. Wikidata: a free collaborative knowledge base

    2014

  19. [27]

    Graph convolutional neural networks for web- scale recommender systems, in: Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp

    Ying, R., He, R., Chen, K., Eksombatchai, P., Hamilton, W.L., Leskovec, J., 2018a. Graph convolutional neural networks for web- scale recommender systems, in: Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp. 974–983

  20. [28]

    Hierarchicalgraphrepresentationlearningwithdifferentiable pooling,in: AdvancesinNeuralInformationProcessingSystems,pp

    Ying, Z., You, J., Morris, C., Ren, X., Hamilton, W., Leskovec, J., 2018b. Hierarchicalgraphrepresentationlearningwithdifferentiable pooling,in: AdvancesinNeuralInformationProcessingSystems,pp. 4800–4810

  21. [29]

    replaces

    Zhang, M., Chen, Y., 2018. Link prediction based on graph neural networks, in: Advances in Neural Information Processing Systems, pp. 5165–5175. Raehyun Kim et al.:Preprint submitted to Elsevier Page 13 of 15 HATS: A Hierarchical Graph Attention Network for Stock Movement Pred...

    2018

  22. [2017]

    Neural message passing for quantum chemistry, in: Pro- ceedings of the 34th International Conference on Machine Learning- Volume 70, JMLR. org. pp. 1263–1272

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.