Pith. sign in

REVIEW 4 major objections 4 minor 40 references

Enhancing Meme Token Market Transparency: A Multi-Dimensional Entity-Linked Address Analysis for Liquidity Risk Evaluation

T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Meme-token liquidity indicators overstate true liquidity, and merging addresses controlled by the same entity corrects the picture.

desk verdict A plausible, clearly written application of address clustering to meme-token liquidity, but the evidentiary core is thinner than the claims and one clustering description is internally inconsistent. read the letter →

arxiv 2506.05359 v1 pith:WEJFPDVB submitted 2025-05-22 q-fin.ST cs.CR

Qiangqiang Liu , Qian Huang , Frank Fan , Haishan Wu , Xueyan Tang This is my paper · ORCID
classification q-fin.STcs.CR
keywords memetokensliquidityriskentity-linkedaddressesaddressclusteringwashtradingconcentrationdecentralizedexchangeblockchainanalysis
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 standard on-chain liquidity figures for meme tokens are systematically too rosy because a single entity can control many wallet addresses. It proposes a multi-dimensional address-identification pipeline that links addresses to the same owner, then recomputes six liquidity indicators on the merged groups. On BabyBonk, NMT, and BonkFork, the adjusted indicators reveal higher ownership concentration and thinner genuine liquidity than the raw numbers suggest, and the adjustment changes which token looks safer to buy. If the method works, token-distribution and trading-activity metrics available on public dashboards systematically understate concentration risk for small-cap tokens.

What carries the argument

The load-bearing object is the entity-linked address group: a set of blockchain addresses whose keys are controlled by the same entity, identified by four complementary heuristics. Source-of-funds analysis links a main wallet to its sub-wallets through one-to-many funding patterns; destination-of-funds analysis links many small wallets that send to one address; behavioral similarity analysis runs Louvain community detection over transaction graphs to catch addresses that act alike without direct fund links; anomalous-transaction analysis flags identical-amount transfers and high-frequency small transfers that signal self-trading or circular trading. The candidate groups are then refined by DBSCAN clustering (Eps=0.5, MinPts=5), isolation forest (contamination 0.1) to drop outliers, and a probabilistic linkage model with a 0.7 threshold, before six indicators — Top-10 share, HHI, VMTV, volume-to-liquidity, DEX pool liquidity, and holder count — are recomputed on the merged groups.

What would settle it

Construct a ground-truth test on a token where a single operator controls, say, 50 fresh addresses funded through a mixer or via an exchange hot wallet with no direct on-chain link to the operator; run the pipeline. If most of those addresses remain unmerged, or if ordinary retail addresses that merely shared an exchange withdrawal are merged, the adjusted Top-10 and HHI would be wrong in opposite directions, and the claimed improvement in risk assessment would not hold.

Watch

Extended reading notes

Core claim

The paper's central claim is that entity-linked address identification exposes a gap between apparent and actual liquidity in meme token markets, and that this gap matters for investment and regulatory decisions. Concretely, it claims that after clustering addresses controlled by the same entity — using one-to-many and many-to-one fund-flow heuristics, Louvain community detection on behavioral similarity, and detection of identical-amount and high-frequency transactions, refined by DBSCAN, isolation forest, and a 0.7 probability threshold — the true top-10 concentration, Herfindahl-Hirschman index, volume-to-market-cap ratio, and volume-to-liquidity ratio change materially. For BabyBonk the adjusted Top-10 position is higher than the raw one, showing that apparent dispersion hides concentration; in the three-token comparison, NMT's adjusted liquidity profile becomes clearly better than BabyBonk's, and NMT subsequently showed relatively higher price stability. The paper presents this as evidence that the framework gives a more accurate liquidity risk assessment than raw transaction data, and as a first comprehensive analytical framework for meme-token liquidity risk.

Load-bearing premise

The whole analysis stands on the assumption that the four heuristics plus DBSCAN, isolation forest, and the 0.7 probability threshold actually separate wallets controlled by one entity from independent retail holders — and that removing labeled contracts, hot wallets, and airdrop addresses leaves enough signal to do so.

Editorial extensions

If this is right

  • Raw Top-10 and HHI figures understate concentration whenever a large holder splits funds across many addresses; the merged figures are the ones that capture rug-pull vulnerability.
  • Self-trading and circular trading inflate VMTV and volatility indicators, so excluding entity-internal trades yields a truer measure of genuine market activity.
  • A token that looks comparable before merging can rank differently afterward: in the March 2024 comparison, NMT's optimized radar area exceeds BabyBonk's, and NMT showed relatively higher price stability subsequently.
  • The identification method is not restricted to BSC; the paper states the core concept transfers to Ethereum, BASE, and Solana with adjusted data preprocessing.

Reading between the lines

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

  • If the pipeline generalizes, any liquidity ranking of small-cap tokens built from raw on-chain data should be read as an upper bound on genuine liquidity, and concentration-corrected rankings could become a standard pre-trade screen.
  • A direct test the paper does not run: take a labeled set of known same-owner addresses (exchange disclosures, confirmed exploit wallets) and measure the precision and recall of the 0.7-threshold linkage; parameter drift across chains is plausible because preprocessing differs by chain.
  • The authors stop short of longitudinal validation; the natural next step implied by their own limitation section is testing whether adjusted indicators predict price crashes or rug-pulls over longer windows.
  • One implicit caveat is that the largest BabyBonk cluster (27.8% of tokens) is inferred to be a market-maker and excluded from risk analysis; whether a group is a market-maker or a manipulator is a judgment call that changes the risk read.
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 / 4 minor

Summary. The paper proposes a multi-dimensional entity-linked address identification method for meme token liquidity risk assessment. The pipeline combines four heuristics (source of funds, destination of funds, behavioral similarity, anomalous transactions) with DBSCAN clustering, isolation forest anomaly detection, and a probabilistic linkage model, and then computes six liquidity indicators (top-10 concentration, HHI, VMTV, volatility, DEX liquidity, holder count) before and after "entity-linked optimization" on three BSC meme tokens (BabyBonk, NMT, BonkFork). The central empirical claim is that apparent liquidity is systematically better than actual liquidity, and that after adjustment NMT is more liquid than BabyBonk.

Significance. If the entity-linkage step were validated, the paper would address a genuine gap: meme tokens are a large and volatile asset class, and standard on-chain concentration metrics ignore that one entity may control many addresses. The paper is also honest about several limitations in Section V, including the lack of longitudinal validation and the static nature of the indicators. However, the empirical claims are not currently reproducible or externally checkable: no code, data, or ground-truth labels are provided, and the clustering step that generates the headline disparities is described only at a high level. The significance is therefore conditional on a substantial validation effort.

major comments (4)
  1. [Section IV.B.5] The paper states that the cluster with label=-1 is "the entity-linked holding 49% of the total tokens" and "has only one address ... indicating many retail investors." In DBSCAN, label=-1 denotes noise points that do not belong to any cluster, not an entity, and a single holder of 49% is the opposite of "many retail investors." This internal inconsistency directly affects the adjusted Top-10 and HHI values, which are the headline evidence for the supposed disparity. Please re-estimate the optimized indicators while treating the 49% holder as a single entity, or clarify what label=-1 means in the post-processing stage.
  2. [Section III.C and Section IV.B.5] The reduction from 18,587 addresses to 5,245 candidate groups and then to 1,214 final entity-linked groups is reported without describing the refinement rules, the group-size distribution, or the merge criteria. Any non-trivial merging mechanically reduces the effective holder count and inflates top-10 and HHI concentration, so the observed "significant disparities" could be an artifact of over-merging rather than evidence of hidden entities. Please report the merging criteria and the distribution of group sizes before and after refinement, and add either precision/recall on held-out labeled addresses or a null-model baseline (for example, random address grouping preserving the same degree distribution).
  3. [Section IV.B and Section V] The evaluation is structurally circular: the same heuristics that define the entity groups are also used as evidence that the groups reveal hidden concentration. Section V acknowledges that identification accuracy may be affected by complex strategies, but it offers no quantification. Please add an external validation layer, such as comparison with Arkham-labeled entities, exchange withdrawal patterns, or a synthetic transaction network with known ground-truth clusters, so that the entity-linkage accuracy can be assessed independently of the heuristics that generated the groups.
  4. [Section IV.C] The comparative claim that NMT is more liquid than BabyBonk after entity-linked optimization rests on three tokens in a single month, a qualitative radar-area comparison, and an anecdotal statement about subsequent price stability. Please report the numerical indicator values before and after optimization for each token, with standard errors or bootstrap confidence intervals, and a sensitivity analysis over the tuning parameters introduced in Section III.C (DBSCAN Epsilon/MinPts, isolation forest contamination rate, and the probabilistic linkage threshold). As written, the cross-token conclusion is not statistically supported.
minor comments (4)
  1. [Figure 1] 'Akrham' should be 'Arkham' in the figure and in the surrounding text.
  2. [Section III.D] The mathematical notation for HHI, VMTV, and Volatility is malformed (for example, 'HHI = nX i=1 p^2_i'); the summation and fraction symbols need to be rendered correctly.
  3. [References] Reference [4] appears in the reference list but is not cited in the text, and in Section II.C the text cites Cumming et al. as [28] twice, although the reference list has [29] for Cumming et al.; please correct the citation numbering.
  4. [Section IV.B.5] The statement that parameters were 'optimized through iterative testing to maximize the accuracy of the entity identification' is not verifiable; please specify the optimization target and the range of parameter values tested, or move this claim to the limitations section.

Circularity Check

1 steps flagged · score 6.0 of 10

Entity-linked 'validation' is structurally circular: the adjusted liquidity indicators are computed from the same unvalidated clusters they are meant to validate.

  1. fitted input called prediction [Section IV.C, MEME Token Liquidity Analysis]
    "We use the identified entity-linked groups to calculate the indicators mentioned in Section 3.3 and compare the results before and after optimization using entity-linked identification. ... We can observe that after optimization using entity-linked groups for BabyBonk, all indicators except the liquidity indicator have decreased. The change in the Top 10 Position is the most significant. After entity-linked identification, we can discover that the actual concentration of the top 10 token holders is higher, revealing the true liquidity risk."

    The 'after optimization' indicators are computed by merging 18,587 addresses into 5,245 and then 1,214 entity-linked groups (Section IV.B.5), where the grouping is produced by the very DBSCAN/Isolation-Forest/probability-threshold pipeline being proposed. The paper reports no ground-truth labels, precision/recall, or null model; Section V concedes the identification accuracy may be affected by complex strategies and that the study lacks longitudinal validation. Therefore the observed decrease in apparent liquidity and increase in top-10 concentration is the arithmetic re-labelling of holders, not independent evidence that raw indicators were biased.

full rationale

The paper does not derive a constant from itself or rely on a self-citation chain, so there is no equation-level self-derivation. The main circularity is evaluative: the empirical 'discovery' of significant disparities between apparent and actual liquidity is produced by the same entity-linked grouping that defines what 'actual' means. No external ground truth, benchmark dataset, or statistical null model is provided, and the clustering parameters are described only as 'optimized through iterative testing to maximize the accuracy' without any accuracy measure. Section V explicitly acknowledges that the accuracy of entity-linked identification may be affected by complex strategies and that the study lacks longitudinal validation, which confirms that the before/after comparison is not an independent test. A further internal inconsistency appears in Section IV.B.5, where the cluster labeled '-1' (noise in DBSCAN) holding 49% of tokens is described as 'indicating many retail investors'; this undermines confidence in the adjusted concentration metrics but is a validity problem rather than a circularity step. Overall, the central validation claim partially reduces by construction because the adjusted liquidity indicators are the direct output of the fitted clustering pipeline, so a score of 6 reflects this partial circularity without alleging full definitional equivalence.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The paper introduces no new physical or mathematical entities. Its free parameters are all hand-set clustering thresholds. Its axioms are standard blockchain-forensics assumptions about fund flow and behavioral similarity, none of which are validated against ground truth in the paper.

free parameters (7)
  • Source of Funds thresholds = minimum 5 receiving addresses, minimum 10 USDT
    Chosen by hand to filter noise in Section IV-B1 and are not derived from data.
  • Destination of Funds thresholds = minimum 5 sending addresses, minimum 10 USDT
    Chosen by hand in Section IV-B2 with no justification beyond filtering noise.
  • Behavioral similarity threshold = minimum 10 USDT
    Chosen by hand in Section IV-B3.
  • Anomalous transaction thresholds = minimum 5 transactions, minimum 5 USDT
    Chosen by hand in Section IV-B4.
  • DBSCAN Eps and MinPts = Eps=0.5, MinPts=5
    Stated as chosen to balance cluster density and noise tolerance in Section IV-B5, with no sensitivity analysis.
  • Isolation forest contamination rate = 0.1
    Stated as based on empirical observations in Section IV-B5.
  • Probabilistic entity linkage threshold = 0.7
    Set conservatively to minimize false positives in Section IV-B5.
assumptions (6)
  • domain assumption Funds originating from the same non-institutional address imply the same entity.
    Stated as a reasonable assumption in Section III-B1.
  • domain assumption Multiple addresses sending to a single address imply the same entity.
    Stated in Section III-B2.
  • domain assumption Louvain community detection on behavioral similarity identifies coordinated entities.
    Invoked in Section III-B3.
  • domain assumption Addresses with identical-amount or high-frequency transfers are entity-linked.
    Stated in Section III-B4.
  • domain assumption Removing contracts, hot wallets, and airdrop participants leaves a graph where remaining edges reflect ownership.
    Appears in Section III-A and is used to define the empirical dataset.
  • domain assumption Radar-chart area and indicator comparisons before and after clustering reflect true liquidity quality.
    Used in Section IV-C to conclude NMT is more liquid than BabyBonk.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Enhancing Meme Token Market Transparency: A Multi-Dimensional Entity-Linked Address Analysis for Liquidity Risk Evaluation." pith.science (2026). https://pith.science/paper/WEJFPDVB

@misc{pith2026250605359,
  author       = {Pith},
  title        = {Pith review of: Enhancing Meme Token Market Transparency: A Multi-Dimensional Entity-Linked Address Analysis for Liquidity Risk Evaluation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WEJFPDVB}},
  note         = {Machine review of arXiv:2506.05359}
}
read the original abstract

Meme tokens represent a distinctive asset class within the cryptocurrency ecosystem, characterized by high community engagement, significant market volatility, and heightened vulnerability to market manipulation. This paper introduces an innovative approach to assessing liquidity risk in meme token markets using entity-linked address identification techniques. We propose a multi-dimensional method integrating fund flow analysis, behavioral similarity, and anomalous transaction detection to identify related addresses. We develop a comprehensive set of liquidity risk indicators tailored for meme tokens, covering token distribution, trading activity, and liquidity metrics. Empirical analysis of tokens like BabyBonk, NMT, and BonkFork validates our approach, revealing significant disparities between apparent and actual liquidity in meme token markets. The findings of this study provide significant empirical evidence for market participants and regulatory authorities, laying a theoretical foundation for building a more transparent and robust meme token ecosystem.

Figures

Figures reproduced from arXiv: 2506.05359 by the authors.

Figure 1
Figure 1. Workflow of the proposed liquidity risk analysis. It comprises the following steps: (A) Gathering data from BscScan, Akrham, and DEXScreener; (B) [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Tree-like fund distribution. 4) Anomalous Transaction Behavior Analysis: By setting thresholds of minimum 5 transactions and 5 USDT as the minimum transaction amount, we filter out noise, resulting in the discovery of 70 entity-linked groups encompassing 2,015 addresses. As shown in the figure below, we selected all transactions with a transfer amount of one hundred billion and plotted a transaction graph for the ad… view at source ↗
Figure 5
Figure 5. Entity-linked groups of cluster id = 0 [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figures from the paper (2 more)
Figure 6
Figure 6. Figure 6: Comparison of BabyBonk token liquidity before and after entity [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: Comparison of BabyBonk/NMT/BonkFork token liquidity before and after entity-linked optimization. [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 39 canonical work pages

  1. [1]

    The Rise of Meme Tokens: A New Paradigm in Cryptocurrency,

    Smith, A., et al., “The Rise of Meme Tokens: A New Paradigm in Cryptocurrency,” Journal of Digital Finance, vol. 15, no. 3, pp. 245- 260, 2022

    work page 2022

  2. [2]

    Meme Tokens and Social Media Influence in Crypto Markets,

    Johnson, L., “Meme Tokens and Social Media Influence in Crypto Markets,” Blockchain Economics Review, vol. 8, no. 2, pp. 112-128, 2023

    work page 2023

  3. [3]

    Limitations of Traditional Financial Analysis in Cryptocur- rency Markets,

    Brown, R., “Limitations of Traditional Financial Analysis in Cryptocur- rency Markets,” Journal of Financial Innovation, vol. 11, no. 4, pp. 567- 582, 2023

    work page 2023

  4. [4]

    Entity Identification Techniques in Blockchain Net- works,

    Zhang, Y ., et al., “Entity Identification Techniques in Blockchain Net- works,” IEEE Transactions on Blockchain Technology, vol. 3, no. 2, pp. 78-95, 2021

    work page 2021

  5. [5]

    Address clustering heuristics for Ethereum,

    Victor, F., “Address clustering heuristics for Ethereum,” in Financial Cryptography and Data Security, Springer, Cham, pp. 617-633, 2020

    work page 2020

  6. [6]

    Understanding Ethereum via graph analysis,

    Chen, T., et al., “Understanding Ethereum via graph analysis,” ACM Transactions on Internet Technology, vol. 20, no. 2, pp. 1-32, 2021

    work page 2021

  7. [7]

    Bitcoin address clustering using graph neural networks,

    Anouar, T., Abdellatif, K., and Taoufik, G., “Bitcoin address clustering using graph neural networks,” in 2023 International Conference on Electrical, Computer, Communications and Mechatronics Engineering (ICECCME), IEEE, pp. 1-6, 2023

    work page 2023

  8. [8]

    Detecting Mix Groups in Cryptocurrency via Dynamic Evolution of Address Clusters,

    Wu, J., et al., “Detecting Mix Groups in Cryptocurrency via Dynamic Evolution of Address Clusters,” in IEEE INFOCOM 2022-IEEE Con- ference on Computer Communications, pp. 2621-2630, 2022

    work page 2022

Show all 40 references
  1. [9]

    A graph embedding approach for detecting Ponzi schemes on Ethereum,

    Camino, R., Torres, C. F., and State, R., “A graph embedding approach for detecting Ponzi schemes on Ethereum,” Expert Systems with Appli- cations, vol. 213, 118879, 2023

    2023

  2. [10]

    Predicting Bitcoin address behavior via deep learning,

    Wang, Y ., Yang, X., Zeng, Y ., and Tang, Y ., “Predicting Bitcoin address behavior via deep learning,” Soft Computing, vol. 26, no. 6, pp. 2809- 2821, 2022

    2022

  3. [11]

    Address-level cryptocurrency classification: A multi-modal fusion framework,

    Zhong, L., Wu, Q., Wang, Z., Wang, S., and Liu, A., “Address-level cryptocurrency classification: A multi-modal fusion framework,” Future Generation Computer Systems, vol. 138, pp. 247-258, 2023

    2023

  4. [12]

    Detecting money laundering in cryptocurrency transactions using machine learning,

    Ostapowicz, M., and ˙Zbikowski, K., “Detecting money laundering in cryptocurrency transactions using machine learning,” Expert Systems with Applications, vol. 213, 118872, 2023

    2023

  5. [13]

    Analyzing hack subnet- works in the bitcoin transaction graph,

    Goldsmith, D., Grauer, K., and Shmalo, Y ., “Analyzing hack subnet- works in the bitcoin transaction graph,” Applied Network Science, vol. 5, no. 1, pp. 1-22, 2020

    2020

  6. [14]

    Systemic liquid- ity risk in decentralized cryptocurrency exchanges,

    Bidler, N., Romagnoli, G., and Battiston, S., “Systemic liquid- ity risk in decentralized cryptocurrency exchanges,” arXiv preprint arXiv:2305.14631, 2023

    2023

  7. [15]

    Liquidity spillovers in cryptocurrency markets,

    Kitzler, J., Liebenberger, H., and Brauneis, A., “Liquidity spillovers in cryptocurrency markets,” Finance Research Letters, 102999, 2023

    2023

  8. [16]

    A novel metric for assessing cryptocurrency liquidity,

    Zheng, W., Huang, W., Xiao, S., and Huang, A., “A novel metric for assessing cryptocurrency liquidity,” Finance Research Letters, vol. 51, 103352, 2023

    2023

  9. [17]

    Liquidity and price discovery in cryptocurrency markets,

    Li, Y ., and Yi, Y ., “Liquidity and price discovery in cryptocurrency markets,” Finance Research Letters, vol. 51, 103386, 2023

    2023

  10. [18]

    Measuring and forecasting liquidity risks in decen- tralized finance,

    Deng, Y ., et al., “Measuring and forecasting liquidity risks in decen- tralized finance,” International Review of Financial Analysis, vol. 84, 102362, 2022

    2022

  11. [19]

    The rich get richer: Dynamics of the cryptocurrency exchange network,

    Kondor, D., G ´abor, I., Iv ´an, G., and Csabai, I., “The rich get richer: Dynamics of the cryptocurrency exchange network,” Scientific reports, vol. 13, no. 1, pp. 1-13, 2023

    2023

  12. [20]

    The initial DEX offering (IDO): A new way of launching cryptocurrency tokens,

    Ante, L., “The initial DEX offering (IDO): A new way of launching cryptocurrency tokens,” Economics of Innovation and New Technology, pp. 1-15, 2022

    2022

  13. [21]

    Intrinsic Value Assessment in Cryptocurrency,

    Patel, K., Seneviratne, S., Gupta, R., and Ganguly, S., “Intrinsic Value Assessment in Cryptocurrency,” arXiv preprint arXiv:2303.14460, 2023

    2023 arXiv

  14. [22]

    A framework for the fundamental analysis of cryptocurrencies,

    Kyriazis, N., Papadamou, S., and Corbet, S., “A framework for the fundamental analysis of cryptocurrencies,” Research in International Business and Finance, vol. 64, 101806, 2023

    2023

  15. [23]

    Pump and dump dynamics in the cryptocurrency market: An updated analysis,

    Kamps, J., and Kleinberg, B., “Pump and dump dynamics in the cryptocurrency market: An updated analysis,” Journal of Empirical Finance, vol. 71, pp. 14-33, 2023

    2023

  16. [24]

    Crypto Wash Trading,

    Cong, L. W., Li, X., Tang, K., and Yang, Y ., “Crypto Wash Trading,” Available at SSRN 4343520, 2023

    2023

  17. [25]

    Detecting and Characterizing Wash Trading in NFT Markets,

    Bian, S., et al., “Detecting and Characterizing Wash Trading in NFT Markets,” arXiv preprint arXiv:2210.05186, 2022

    2022 arXiv

  18. [26]

    Wash Trading Detection in Cryptocurrency Markets: A Graph-Based Approach,

    Grasso, A., Lagana, D., Piccolo, S., and Scibona, F., “Wash Trading Detection in Cryptocurrency Markets: A Graph-Based Approach,” arXiv preprint arXiv:2303.12626, 2023

    2023 arXiv

  19. [27]

    Meme Digital Assets: Regulatory Challenges and Opportunities,

    Chohan, U. W., “Meme Digital Assets: Regulatory Challenges and Opportunities,” in Regulating Blockchain & Crypto Assets, Springer, Singapore, 2023

    2023

  20. [28]

    Decentralized Finance (DeFi) Policy-Maker Toolkit,

    Zetzsche, D. A., Arner, D. W., and Buckley, R. P., “Decentralized Finance (DeFi) Policy-Maker Toolkit,” University of Hong Kong Faculty of Law Research Paper, no. 2023/03, 2023

    2023

  21. [29]

    Regulation of global crypto exchanges,

    Cumming, D. J., Johan, S., and Pant, A., “Regulation of global crypto exchanges,” Journal of International Business Policy, pp. 1-20, 2023

    2023

  22. [30]

    Cross-chain non-fungible token markets: A data science approach,

    Ante, L., and Meyer, A., “Cross-chain non-fungible token markets: A data science approach,” Finance Research Letters, vol. 49, 103083, 2022

    2022

  23. [31]

    Explosive bubbles in the cryptocurrency market: New evidence from thousands of altcoins,

    Shanaev, S., Shuraeva, A., Sharma, S., and Ghimire, B., “Explosive bubbles in the cryptocurrency market: New evidence from thousands of altcoins,” Finance Research Letters, vol. 51, 103252, 2023

    2023

  24. [32]

    The destabilising effects of cryptocurrency cyber- criminality,

    Corbet, S., et al., “The destabilising effects of cryptocurrency cyber- criminality,” Economics Letters, vol. 222, 110891, 2023

    2023

  25. [33]

    Cryptocurrency Price Prediction Based on Graph Neural Networks,

    Liu, M., et al., “Cryptocurrency Price Prediction Based on Graph Neural Networks,” Mathematics, vol. 11, no. 2, 470, 2023

    2023

  26. [34]

    Transaction- graph based anomaly detection in blockchain networks: Current state-of- the-art and open issues,

    Zheng, Z., Dai, H. N., Chen, W., Wang, H., and Imran, M., “Transaction- graph based anomaly detection in blockchain networks: Current state-of- the-art and open issues,” Journal of Network and Computer Applications, vol. 207, 103552, 2023

    2023

  27. [35]

    Predicting cryptocurrency returns using transfer learning with graph neural networks,

    Huang, T. C., Chen, Y . Y ., and Lai, J. Y ., “Predicting cryptocurrency returns using transfer learning with graph neural networks,” Expert Systems with Applications, vol. 212, 118715, 2023

    2023

  28. [36]

    From Louvain to Leiden: guaranteeing well-connected communities,

    Traag, V . A., Waltman, L., and Van Eck, N. J., “From Louvain to Leiden: guaranteeing well-connected communities,” Scientific Reports, vol. 9, no. 1, 5233, 2019

    2019

  29. [37]

    DBSCAN revisited, revisited: Why and how you should (still) use DBSCAN,

    Schubert, E., Sander, J., Ester, M., Kriegel, H. P., and Xu, X., “DBSCAN revisited, revisited: Why and how you should (still) use DBSCAN,” ACM Transactions on Database Systems (TODS), vol. 42, no. 3, pp. 1-21, 2017

    2017

  30. [38]

    Isolation-based anomaly detection,

    Liu, F. T., Ting, K. M., and Zhou, Z. H., “Isolation-based anomaly detection,” ACM Transactions on Knowledge Discovery from Data (TKDD), vol. 6, no. 1, pp. 1-39, 2012

    2012

  31. [39]

    Naldi, M., and Flamini, M., ”The CR4 index and the interval estimation of the Herfindahl-Hirschman Index: an empirical comparison,” HAL Open Science, hal-01288714, 2014

    2014

  32. [40]

    86, 102320, 2023

    Liang, X., and Wei, J., ”Liquidity risk and expected cryptocurrency returns,” International Review of Financial Analysis, vol. 86, 102320, 2023

    2023

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.