Pith. sign in

REVIEW 4 major objections 8 minor 28 references

Efficient Bitcoin Address Classification Using Quantum-Inspired Feature Selection

T0 review · 4 major / 8 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read 23 quantum-selected features cut Bitcoin mixer training time 30%

desk verdict The 30% training-time saving is mechanical, but the 'maintaining 91% F1' claim is compromised by feature selection before cross-validation; worth a serious revision, not a desk reject. read the letter →

arxiv 2411.15425 v1 pith:SXPYBD3D submitted 2024-11-23 quant-ph cs.CR

Ming-Fong Sie , Yen-Jui Chang , Chien-Lung Lin , Ching-Ray Chang , Shih-Wei Liao This is my paper · ORCID
classification quant-phcs.CR
keywords BitcoinmixeraddressesfeatureselectionQUBOsimulatedannealingquantumrandomforestblockchainanalytics
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 tries to establish that a quantum-inspired feature-selection step can shrink a Bitcoin transaction dataset to the features that matter for identifying mixer addresses. The method casts feature selection as a binary optimization problem that rewards features correlated with the mixer label and penalizes redundant features, and it solves that problem with simulated annealing and quantum annealing. On 694,676 labeled addresses, simulated annealing picked 23 of 69 features; the resulting random forest trained about 30% faster and kept a 91% F1-score for mixers, versus 92% with all features. If correct, the practical payoff is cheaper, faster screening of high-risk Bitcoin addresses.

What carries the argument

The central object is a QUBO cost function with binary variables $x_j$ marking whether feature $j$ is selected: $$f(x) = -\$\alpha$ \sum_{j=1}^n x_j |\rho_{oj}| + (1-\$\alpha$)\sum_{j=1}^n\sum_{k\neq j} x_j x_k |\rho_{jk}|,$$ where $|\rho_{oj}|$ is the Spearman rank correlation between feature $j$ and the binarized mixer label and $|\rho_{jk}|$ is the Spearman correlation between features $j$ and $k$. The first term rewards predictive influence and the second penalizes redundancy, with $\alpha$ balancing the two. The paper minimizes this objective with simulated annealing, quantum annealing, and a hybrid BQM solver, then uses the selected variables as the feature set for supervised classifiers, chiefly random forest.

What would settle it

Repeat the experiment with the feature-selection step redone inside each of the ten training folds, then compare the mixer F1 score to 0.91. If the nested F1 falls well below 0.91, the reported gain partly comes from seeing test labels during selection.

Watch

Extended reading notes

Core claim

The paper's central claim is that a quadratic unconstrained binary optimization (QUBO) feature selector, solved with simulated annealing, can identify the transaction-history features most predictive of Bitcoin mixer use and thereby make a random forest classifier substantially cheaper to train without losing meaningful accuracy. On a dataset of 694,676 labeled addresses with 69 summarized features each, the simulated-annealing selector chose 23 features; the resulting SA Random Forest reached a mixer F1-score of 0.91 and AUC of 0.99 while training in 212 minutes, compared with F1 0.92, AUC 0.99, and 305 minutes for the random forest using all features. Quantum annealing and a hybrid binary-quadratic-model solver selected only 9 and 7 features and produced weaker classifiers (F1 0.60 and 0.70), so the reported practical win is specifically the simulated-annealing route. The authors interpret this as evidence that quantum-inspired combinatorial feature selection can accelerate blockchain analytics on high-risk address detection.

Load-bearing premise

The pipeline's advertised F1 depends on the feature selector never seeing the evaluation labels: the paper computes Spearman correlations and solves the QUBO before the 10-fold classifier split, so the honest comparison requires the selection to be repeated inside each fold.

Editorial extensions

If this is right

  • Analysts can train a random forest on 23 rather than 69 transaction-history features and still detect mixer addresses with F1 around 0.91, recall 0.97, and AUC 0.99.
  • Using all features remains slightly more accurate (F1 0.92), so the practical trade-off is about 1 point of F1 for a 30% reduction in training time.
  • Simulated annealing is the effective optimizer in this setup: quantum annealing selected fewer features but classifiers trained on them reached only F1 0.60 to 0.75.
  • The selected features center on transaction counts, balances, lifetime, and higher-order moments, with lifetime emerging in the SA-selected model.
  • The authors propose the method as transferable to other high-dimensional data domains, including cybersecurity, healthcare, and predictive maintenance.

Reading between the lines

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

  • A test the paper does not run: replacing the QUBO solver with a simple top-$k$ Spearman ranking of features would reveal how much of the 30% training-time saving comes from the redundancy term rather than from dimensionality reduction alone.
  • Because the selection appears to use the full dataset before cross-validation, the 0.91 F1 is likely to drop under a nested selection protocol; the size of the drop would quantify the leakage.
  • The same QUBO-plus-simulated-annealing recipe could be applied to other rare-address classes such as gambling or faucets, where the paper reports weak F1, to see whether feature selection helps or hurts minority classes.
  • For larger feature sets, the 0.0055-second SA timing may not scale linearly; the practical speedup claim should be re-measured as the number of candidate features grows.
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 / 8 minor

Summary. The paper proposes a quantum-inspired feature selection (QIFS) method based on a QUBO objective optimized by simulated annealing (SA), D-Wave quantum annealing (QA), and a hybrid BQM solver, applied to Bitcoin address classification with a focus on detecting mixer addresses. Using 69 transaction-history features and seven classifiers, the authors report that SA-selected features (23 features) reduce random forest training time by 30.3% while maintaining an F1-score of 0.91 for the mixer class, compared with 0.92 using all features. The paper also reports performance for QA and QA BQM feature subsets and analyzes feature importance and class-wise metrics.

Significance. If the central claim were established, the contribution would be a useful empirical demonstration that QUBO-based feature selection can lower computational cost in blockchain analytics with minimal loss in detection quality. The paper also has strengths: it reports a substantial original data-collection effort, makes code available, and the arithmetic behind the 30.3% training-time reduction is consistent with the timing entries in Table 6. However, the evaluation protocol suffers from a selection-leakage problem that directly affects the headline F1/AUC numbers, and the absence of any baseline feature selector means the reported improvement cannot currently be attributed to the proposed method. The significance is therefore conditional on a corrected re-analysis.

major comments (4)
  1. [Sections 3.4 and 3.5] Feature selection is performed on the full dataset before the 10-fold cross-validation. Section 3.4 describes binarizing the classes and computing Spearman correlations between each feature and the mixer label over the whole dataset, then solving the QUBO to select features. Section 3.5 then applies 10-fold cross-validation only to classifier training and hyperparameter optimization. On this description, the held-out test-fold labels influence which features are selected, so the F1 and AUC values in Tables 6 and 7 are optimistically biased. The paper never states that feature selection is nested inside each training fold or performed on a separate selection set. Please re-run the evaluation with nested cross-validation (or a strict train/validation/test split) and report whether the 0.91 F1 and 0.99 AUC for SA Random Forest survive.
  2. [Tables 6 and 7] There is no comparison against simple feature-selection baselines. The central claim is that QIFS with SA is valuable, but the only comparison is between 23 SA-selected features and all 69 features. A 30.3% training-time reduction is largely the mechanical consequence of using fewer features; whether the selected features are good ones cannot be assessed without baselines such as random 23-feature subsets, the top-23 features by absolute Spearman correlation, or classical selectors like mRMR or chi-square. Without such baselines, the paper does not show that the QUBO/SA procedure, rather than merely reducing dimensionality, drives the reported result.
  3. [Tables 6 and 7] Several QA and QA BQM rows are numerically identical despite the two methods selecting different feature sets. In Table 7, QA Random Forest and QA BQM Random Forest have the same AUC vector (0.79, 0.80, 0.78, 0.88, 0.93, 0.85); the same holds for XGBoost, LightGBM, SVM *10%, and Neural Network *10%. In Table 6, QA XGBoost and QA BQM XGBoost are identical in precision, recall, F1, accuracy, AUC, and training time, as are QA LightGBM and QA BQM LightGBM. If these entries are not typographical errors, the results contradict the different feature sets; if they are errors, the tables need correction. This issue must be resolved before the comparative claims about QA versus QA BQM can be evaluated.
  4. [Section 4 and Table 6] The paper reports F1 0.91 for SA Random Forest versus 0.92 for full-feature Random Forest but gives no variance, confidence intervals, or per-fold results. Given the class distribution in Table 1 and the likely correlation of metrics across folds, a 0.01 F1 difference may be within noise. The authors should report fold-level statistics or at least standard deviations for the key comparisons, especially because the headline claim is that accuracy is 'maintained' rather than degraded.
minor comments (8)
  1. [Abstract and Section 2.2] The abstract calls both SA and QA 'quantum-inspired,' but quantum annealing on D-Wave hardware is actual quantum computing, not quantum-inspired. Please use consistent terminology.
  2. [Section 3.3] The text says the features are categorized into five groups, but Table 2 lists four groups: Basic Statistics, Extra Statistics, Moments, and Transaction Patterns. Please correct the count or add the missing group.
  3. [Section 3.4] The GitHub URL 'https://github.com/Siemingfong/Quantom Annealing' contains a space and a typo ('Quantom'); it should be a valid URL, e.g., with '%20' or a corrected repository name.
  4. [Section 3.5] The sentence 'We training of seven common classification algorithms' is ungrammatical; please revise.
  5. [Section 3.5 and Table 6] The text says 'SA Random Forest exhibits a faster 30.3% training time 212 minutes and 23.2 seconds compared to Quantum Random Forest 305 minutes and 5.7 seconds.' The phrase 'Quantum Random Forest' should presumably be 'Random Forest,' and the sentence should be rewritten for clarity.
  6. [Table 7] The table title 'Evaluation of All Features and Full Features and SA QUBO Feature Selection' is redundant ('All Features' and 'Full Features' are the same) and does not mention QA/QA BQM rows. Please revise.
  7. [Section 4 and Figures 6-9] Precision and recall values are reported as percentages with a percent sign in some places and as decimals in others (e.g., '0.91%' versus '0.91'); please standardize the notation.
  8. [Section 5.1] The future-work section states that data imbalance is a limitation and proposes SMOTE, but Table 1 shows that the Mixer class is the largest class (289,006 addresses). If the focus is mixer detection, the class is not a minority; please clarify which classes are imbalanced and how this affects the mixer-specific results.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the QUBO feature-selection objective is explicit and the classifier evaluation is empirical, so no reported number reduces to its inputs by construction.

full rationale

The paper's derivation chain is not circular under the operative definition. The QUBO objective in Section 2.1 maximizes the absolute Spearman correlation between each candidate feature and the binarized mixer label (the |rho_oj| term), which is a stated feature-selection criterion, not a disguised definition of the evaluation metric. The selected feature subset is then used to train classifiers, and the reported F1, AUC, and training times in Tables 6 and 7 are measured outcomes of that training, not algebraic consequences of the QUBO objective. The 30.3% training-time comparison is mechanical in the sense that the SA model uses 23 features instead of 69, but the paper does not claim this reduction is predicted from the objective; it is an empirical measurement. Self-citations to Chang et al. [2,10] and Lin et al. [26] are used for background and for provenance of the hand-crafted feature families, and they are not load-bearing for the SA-Random Forest versus Random Forest comparison, which is computed within this paper. One genuine methodological concern is that Section 3.4 computes Spearman correlations and solves the QUBO on the full dataset before the 10-fold cross-validation described in Section 3.5, so the test folds may influence feature selection. That is a selection-leakage bias and should be addressed by nested cross-validation, but it is not circularity: the reported F1 is not forced by construction, and correcting the protocol would change the estimated magnitude, not the definitional independence of the comparison. Therefore no circular step is identified and the score is 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the accuracy of the WalletExplorer labels, the representativeness of the truncated transaction histories, the choice of Spearman correlation as the relevance measure, and the unstated QUBO weight alpha and annealing settings. No new physical entities are introduced.

free parameters (3)
  • QUBO weight alpha
    Weights the feature influence term against the redundancy term in the objective in Section 2.1. The paper never states its value or how it was chosen.
  • Selected feature count K = 23 for SA, 9 for QA, 7 for QA BQM
    The number of selected features is not constrained in the QUBO; each optimizer returns a different count. Training time comparisons therefore mix the effect of feature count with the effect of which features were chosen.
  • Simulated annealing hyperparameters
    Annealing schedule, number of reads, and beta range for dwave-neal are not reported, so the SA result is not fully specified.
assumptions (4)
  • domain assumption WalletExplorer labels are accurate ground truth for address categories.
    All training and evaluation depends on these labels (Section 3.2). No validation of label quality is provided.
  • domain assumption The first 1,000 transactions per address are representative of the address's behavior.
    The paper states it analyzed only the first 1,000 transactions per address (Section 3.1); high-volume mixers may be mischaracterized.
  • domain assumption Spearman rank correlation with the binary mixer label is a sufficient relevance signal for tree-based classifiers.
    The QUBO objective in Section 2.1 uses |rho_oj| as the influence term; no evidence is given that this correlates with information gain in random forests.
  • ad hoc to paper The QUBO feature selection model of Milne et al. transfers to blockchain transaction features.
    The objective is adopted without modification from a commercial white paper and its effectiveness on this feature set is not established beyond the reported experiments.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Efficient Bitcoin Address Classification Using Quantum-Inspired Feature Selection." pith.science (2026). https://pith.science/paper/SXPYBD3D

@misc{pith2026241115425,
  author       = {Pith},
  title        = {Pith review of: Efficient Bitcoin Address Classification Using Quantum-Inspired Feature Selection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SXPYBD3D}},
  note         = {Machine review of arXiv:2411.15425}
}
read the original abstract

Over 900 million Bitcoin transactions have been recorded, posing considerable challenges for machine learning in terms of computation time and maintaining prediction accuracy. We propose an innovative approach using quantum-inspired algorithms implemented with Simulated Annealing and Quantum Annealing to address the challenge of local minima in solution spaces. This method efficiently identifies key features linked to mixer addresses, significantly reducing model training time. By categorizing Bitcoin addresses into six classes: exchanges, faucets, gambling, marketplaces, mixers, and mining pools, and applying supervised learning methods, our results demonstrate that feature selection with SA reduced training time by 30.3% compared to using all features in a random forest model while maintaining a 91% F1-score for mixer addresses. This highlights the potential of quantum-inspired algorithms to swiftly and accurately identify high-risk Bitcoin addresses based on transaction features.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 21 canonical work pages

  1. [1]

    Reviews of Modern Physics 39, 883–893 (1967)

    Brush, S.G.: History of the lenz-ising model. Reviews of Modern Physics 39, 883–893 (1967)

    work page 1967

  2. [2]

    IEEE Nanotechnology Magazine 18(3), 15–22 (2024) https://doi.org/10.1109/MNANO.2024.3378485

    Chang, Y.-J., Nien, C.-F., Huang, K.-P., Zhang, Y.-T., Cho, C.-H., Chang, C.-R.: Quantum computing for optimization with ising machine. IEEE Nanotechnology Magazine 18(3), 15–22 (2024) https://doi.org/10.1109/MNANO.2024.3378485

    arXiv 2024

  3. [3]

    In: Data Management, Analytics and Innovation: Proceedings of ICDMAI 2020, Volume 1, pp

    Mandal, A.K., Panday, M., Biswas, A., Goswami, S., Chakrabarti, A., Chakraborty, B.: An approach of feature subset selection using simulated quan- tum annealing. In: Data Management, Analytics and Innovation: Proceedings of ICDMAI 2020, Volume 1, pp. 133–146 (2021). Springer

    work page 2021

  4. [4]

    Decentralized business review (2008) 19

    Nakamoto, S.: Bitcoin: A peer-to-peer electronic cash system. Decentralized business review (2008) 19

    work page 2008

  5. [5]

    Empirical Economics 61(5), 2663–2683 (2021)

    Baur, D.G., Dimpfl, T.: The volatility of bitcoin and its role as a medium of exchange and a store of value. Empirical Economics 61(5), 2663–2683 (2021)

    work page 2021

  6. [6]

    In: 2022 IEEE International Conference on Big Data (Big Data), pp

    Rathore, M.M., Chaurasia, S., Shukla, D.: Mixers detection in bitcoin network: a step towards detecting money laundering in crypto-currencies. In: 2022 IEEE International Conference on Big Data (Big Data), pp. 5775–5782 (2022). IEEE

    work page 2022

  7. [7]

    Frontiers in Physics 9, 665399 (2021)

    Liu, M., Chen, H., Yan, J.: Detecting roles of money laundering in bitcoin mixing transactions: A goal modeling and mining framework. Frontiers in Physics 9, 665399 (2021)

    work page 2021

  8. [8]

    Sensors 22(19), 7162 (2022)

    Ashfaq, T., Khalid, R., Yahaya, A.S., Aslam, S., Azar, A.T., Alsafari, S., Hameed, I.A.: A machine learning and blockchain based efficient fraud detection mechanism. Sensors 22(19), 7162 (2022)

    work page 2022

Show all 28 references
  1. [9]

    In: Proceedings of the Web Conference 2021, pp

    Wu, L., Hu, Y., Zhou, Y., Wang, H., Luo, X., Wang, Z., Zhang, F., Ren, K.: Towards understanding and demystifying bitcoin mixing services. In: Proceedings of the Web Conference 2021, pp. 33–44 (2021)

    2021

  2. [10]

    IEEE Nanotechnology Magazine 17(2), 31–37 (2023) https://doi.org/10.1109/MNANO.2023.3249501

    Chang, Y.-J., Sie, M.-F., Liao, S.-W., Chang, C.-R.: The prospects of quantum computing for quantitative finance and beyond. IEEE Nanotechnology Magazine 17(2), 31–37 (2023) https://doi.org/10.1109/MNANO.2023.3249501

    2023

  3. [11]

    Machine learning 45, 5–32 (2001)

    Breiman, L.: Random forests. Machine learning 45, 5–32 (2001)

    2001

  4. [12]

    In: Proceedings of the 22nd Acm Sigkdd International Conference on Knowledge Discovery and Data Mining, pp

    Chen, T., Guestrin, C.: Xgboost: A scalable tree boosting system. In: Proceedings of the 22nd Acm Sigkdd International Conference on Knowledge Discovery and Data Mining, pp. 785–794 (2016)

    2016

  5. [13]

    Clinical chemistry 39(4), 561–577 (1993)

    Zweig, M.H., Campbell, G.: Receiver-operating characteristic (roc) plots: a fun- damental evaluation tool in clinical medicine. Clinical chemistry 39(4), 561–577 (1993)

    1993

  6. [14]

    https://1qbit.com

    Milne, A., Rounds, M., Goddard, P.: Optimal Feature Selection in Credit Scoring and Classification Using a Quantum Annealer. https://1qbit.com. White Paper, 1QB Information Technologies (2017)

    2017

  7. [15]

    Reports on Progress in Physics 83 (2019) https://doi.org/10.1088/1361-6633/ab85b8

    Hauke, P., Katzgraber, H., Lechner, W., Nishimori, H., Oliver, W.: Perspectives of quantum annealing: methods and implementations. Reports on Progress in Physics 83 (2019) https://doi.org/10.1088/1361-6633/ab85b8

    2019 doi

  8. [16]

    Science 220(4598), 671–680 (1983)

    Kirkpatrick, S., Gelatt Jr, C.D., Vecchi, M.P.: Optimization by simulated annealing. Science 220(4598), 671–680 (1983)

    1983

  9. [17]

    https://pypi.org/project/dwave-neal/

    Inc., D.-W.S.: dwave-neal: A Python Simulated Annealing Sampler Library (2023). https://pypi.org/project/dwave-neal/

    2023

  10. [18]

    Circulation 117(18), 2395–2399 (2008) 20

    LaValley, M.P.: Logistic regression. Circulation 117(18), 2395–2399 (2008) 20

    2008

  11. [19]

    Journal of computer and system sciences 55(1), 119–139 (1997)

    Freund, Y., Schapire, R.E.: A decision-theoretic generalization of on-line learning and an application to boosting. Journal of computer and system sciences 55(1), 119–139 (1997)

    1997

  12. [20]

    Statistics and its Interface 2(3), 349–360 (2009)

    Hastie, T., Rosset, S., Zhu, J., Zou, H.: Multi-class adaboost. Statistics and its Interface 2(3), 349–360 (2009)

    2009

  13. [21]

    In: Advances in Neural Information Processing Systems, pp

    Ke, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., Ye, Q., Liu, T.- Y.: Lightgbm: A highly efficient gradient boosting decision tree. In: Advances in Neural Information Processing Systems, pp. 3146–3154 (2017)

    2017

  14. [22]

    Machine learning 20, 273–297 (1995)

    Cortes, C., Vapnik, V.: Support-vector networks. Machine learning 20, 273–297 (1995)

    1995

  15. [23]

    the Journal of machine Learning research 12, 2825–2830 (2011)

    Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V.,et al.: Scikit-learn: Machine learning in python. the Journal of machine Learning research 12, 2825–2830 (2011)

    2011

  16. [24]

    In: Advances in Neural Information Processing Systems, pp

    Louppe, G., Wehenkel, L., Sutera, A., Geurts, P.: Understanding variable impor- tances in forests of randomized trees. In: Advances in Neural Information Processing Systems, pp. 431–439 (2013)

    2013

  17. [25]

    Toyoda, K., Ohtsuki, T., Mathiopoulos, P.T.: Multi-class bitcoin-enabled service identification based on transaction history summarization. In: 2018 IEEE Inter- national Conference on Internet of Things (iThings) and IEEE Green Computing and Communications (GreenCom) and IEEE ...

    2018

  18. [26]

    In: 2019 IEEE International Conference on Blockchain and Cryptocurrency (ICBC), pp

    Lin, Y.-J., Wu, P.-W., Hsu, C.-H., Tu, I.-P., Liao, S.-w.: An evaluation of bitcoin address classification based on transaction history summarization. In: 2019 IEEE International Conference on Blockchain and Cryptocurrency (ICBC), pp. 302–310 (2019). IEEE

    2019

  19. [27]

    Journal of artificial intelligence research 16, 321–357 (2002)

    Chawla, N.V., Bowyer, K.W., Hall, L.O., Kegelmeyer, W.P.: Smote: synthetic minority over-sampling technique. Journal of artificial intelligence research 16, 321–357 (2002)

    2002

  20. [28]

    Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery 1 (2011) https://doi.org/10

    Seliya, N., Khoshgoftaar, T.: The use of decision trees for cost-sensitive classifi- cation: an empirical study in software quality prediction. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery 1 (2011) https://doi.org/10. 1002/widm.38 21

    2011

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.