Pith. sign in

REVIEW 3 major objections 8 minor 58 references

Financial Audit Assistance using Misinformation Detection and Explanation

T0 review · 3 major / 8 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 An unsupervised pipeline that mines past audit reports can flag financial statements likely to contain misinformation and identify the financial variables responsible.

desk verdict Useful integration of the authors' own audit-assistance pipeline with a real new T5 classifier, but the headline numbers rest on self-generated silver labels and an explanation evaluation that drops 13 of 20 flagged companies. read the letter →

arxiv 2607.17797 v1 pith:4TH2GD4F submitted 2026-07-20 cs.AI

Kshitij Madhav Jadhav , Sushodhan Vaishampayan , Manoj Apte , Sachin Pawar , Nitin Ramrakhiyani , Girish Keshav Palshikar This is my paper · ORCID
classification cs.AI
keywords financialauditingmisinformationdetectionstatementsanomalyexplainableAIauditreportsXBRLsilverlabels
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 claims that a fully unsupervised pipeline can scan a large corpus of financial statements, flag the company-years most likely to contain misinformation, and point to the specific financial variables that explain the flag. The system builds its own training signal from past audit reports: sentences that two independent language models agree are adverse remarks become silver labels for the matching statements. It then applies anomaly-detection and regression models to 11,460 statements over five years, with SVR regression reaching an average precision-at-20 of 0.83, and explains the flags with methods whose best average F1 against silver explanations is 0.24. If the claims hold, auditors could use the system as a scalable pre-audit triage tool that reduces manual effort and subjectivity.

What carries the argument

The load-bearing machinery is the silver-label loop: an audit sentence is 'adverse' only if a fine-tuned text-to-text classifier and a large language model both say so, and a company-year is 'misinformation' only if at least two such sentences appear in its audit report (current or following year). This loop turns unstructured audit prose into the labels used to score every detector. The detectors themselves are step-wise regression models over liability-related variables (with SVR the strongest) and a battery of anomaly-detection algorithms. Explanations are produced by three outlying-aspect methods—Mahalanobis-rank subsets (EMD), Isolation-Forest path summaries (EiForest), and maximal-isolation integer-programming rules (EMI)—and the silver-standard explanations they are measured against are built by matching adverse sentences to XBRL taxonomy categories and mapping those up to the 42 financial variables.

What would settle it

Run the same SVR and explanation methods on a corpus where the silver labels are replaced by confirmed accounting restatements or regulatory enforcement findings; if the average precision-at-20 drops to near random, the claim that the system detects misinformation would be refuted.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that misinformation in financial statements leaves detectable statistical traces in the structured numbers, and those traces can be found without any manually labeled examples. Past statements and their audit reports provide the supervision: a statement is silver-labeled as containing misinformation when its audit report has at least two sentences that both language models classify as adverse remarks, using either the current or the following year's report. Against these labels, step-wise regression models built for liability-related variables, especially SVR, identify the top 20 suspect statements per year with average precision up to 0.83, outperforming 18 anomaly-detection algorithms (best average 0.56). For explanation, the EiForest method, which summarizes how a random-partition forest isolates a point, achieves the best average F1 of 0.24 when matched against silver-standard explanations constructed by mapping adverse-remark sentences to XBRL categories and then to the 42 financial variables.

Load-bearing premise

The evaluation assumes that a company-year truly contains misinformation whenever its audit report has at least two sentences that both language models label as adverse remarks, so the precision numbers are only as good as that silver-label rule.

Editorial extensions

If this is right

  • An audit team could run the pipeline over thousands of unlabeled filings and spend detailed review hours on the top-20 list per year.
  • The silver-labeling procedure can label any corpus of statements that comes with audit reports, removing the need for manual annotation.
  • Using the following year's audit report lets the system catch misinformation whose effects accumulate and only become visible later.
  • The explanation output gives auditors a concrete starting point, naming specific balance-sheet, income-statement, or cash-flow variables to check.
  • Because every step from labeling to explanation is automatic, the pre-audit triage becomes scalable and less subjective.

Reading between the lines

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

  • The silver labels are the weakest link: if the two-model agreement threshold does not track actual restatements or fraud, the reported precision numbers overstate real-world performance; replacing silver labels with confirmed restatement events would test this directly.
  • An F1 of 0.24 means auditors should treat variable-level explanations as hypotheses to verify, not as evidence, even when the detection ranking itself is useful.
  • The pipeline should transfer to other jurisdictions or filing regimes if the XBRL taxonomy and audit-report language are available, but the paper's single-country, five-year corpus leaves that transfer untested.
  • Precision is only measured at the top 20 of each year; the paper does not report how detection degrades deeper into the ranked list, which is what a full-population screening would need to know.
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 / 8 minor

Summary. The paper describes an integrated, unsupervised pipeline for financial-audit assistance. On 11,460 company-year financial statements (2,292 Indian listed companies, 2010-2014) with associated audit reports, it builds silver-standard misinformation labels by requiring at least two sentences in the current or next year audit report that both a fine-tuned T5 classifier and Mistral-7B-Instruct label as adverse remarks. It then tests unsupervised anomaly detectors and step-wise regression models, reporting precision@20 against those labels (best average P@20 = 0.83 for SVR M17, LOF 0.56 among anomaly detectors). For flagged companies it generates explanations via EMD, EiForest, and EMI, and compares them with SHAP and LIME against silver-standard explanations built by mapping adverse sentences to XBRL categories and then to financial variables. The best explanation result is EiForest with average F1 = 0.24, computed on only 7 of 20 flagged companies. The paper concludes that the system detects FS likely to contain misinformation with good precision and that the explanation methods identify the responsible variables.

Significance. If the detection and explanation metrics reflected confirmed misinformation, the system would be a useful scalable pre-audit triage tool. The paper's strengths include a large multi-year corpus, a fully unsupervised detection pipeline, manual annotation of two adverse-remark sentence sets with reported inter-annotator agreement, and evaluation of the T5 sentence classifier on held-out data. However, because the central metrics are measured against the authors' own model-derived silver labels and the explanation evaluation is restricted to a favorable subset, the current evidence does not support the absolute "good precision" and explanation-accuracy claims. The work is potentially valuable as a demonstration of an end-to-end auditor-assistance pipeline, but the evaluation needs independent anchoring or a substantial reframing.

major comments (3)
  1. [Section 3, dataset D(SL)_str, and Section 4] The load-bearing detection metric is evaluated against silver labels created by the authors' own pipeline: a company-year is labeled MISINFO only if at least two sentences are labeled adverse by both a T5 model trained on automatically labeled sentences and Mistral-7B-Instruct, and the positive condition can also be satisfied by the next year's audit report. Consequently, P@20 in Tables 6 and 8 measures consistency with this two-model agreement criterion rather than the presence of confirmed misinformation. Since no external anchor (e.g., subsequent restatements, regulatory actions, or a manually verified sample) is used, the conclusion that the system detects FS "likely containing misinformation with good precision" is not calibrated to real-world misinformation. Please add an independent validation subset or explicitly reframe the claims as agreement with the model-defined criterion.
  2. [Section 5.6, Tables 11-12] The explanation evaluation is selectively reported: of the top 20 companies flagged by SVR M15 in 2014, only 7 appear in Tables 11-12; the other 13 are excluded because the T5/Mistral adverse-sentence intersection is empty or because the intersection of all generated explanations with the silver explanation is empty. These are system failures on 13 of 20 flags, not missing data. If the 13 excluded companies are scored as F1 = 0, EiForest's average F1 falls from 0.24 to about 0.08, and the claim that explanations identify the responsible variables no longer holds for the full flagged set. The paper should report metrics on all 20 companies (including zero scores) or define a different, justified evaluation protocol.
  3. [Section 5.5 and Section 5.6] The silver-standard explanations are constructed from the same adverse-remark sentences used to create the silver misinformation labels: adverse sentences are mapped to XBRL categories and then to financial variables, and the union of those variables is treated as ground truth. This creates a circularity in the explanation evaluation: an explanation method that happens to name variables similar to those in the audit report's adverse sentences will score well by construction, regardless of whether those variables are the actual source of misinformation in the FS. An evaluation against independently constructed explanations, or against restatement data, is needed before attributing explanatory power to the methods.
minor comments (8)
  1. [Sections 1 and 3] "Web-scrapped" should be "web-scraped" in both places where it appears.
  2. [Section 4.1] "Descent average precision" should be "decent average precision."
  3. [Section 6] The sentence "showed that they they detect companies" contains a duplicated "they" and should read "showed that they detect companies."
  4. [Section 5.6] "We use the 2024 subset of the silver label dataset" is inconsistent with the 2010-2014 data and should read "2014 subset."
  5. [Section 5.6] The sentence beginning "To cover such companies, We could relax our criterion" is a dangling suggestion; either implement this relaxation or remove the sentence.
  6. [References] Reference [56] cites Kuhn and Tucker, but the SHAP method should be cited to Lundberg and Lee (NeurIPS 2017).
  7. [Table 1 and Section 3] The notation for the silver dataset is inconsistent: D(SL)_str in Table 1 versus D_SL_str in the text; please unify.
  8. [Table 11] The caption "Explanations generated by all the methods" is misleading because the table also includes the ground-truth column; please state this explicitly.

Circularity Check

3 steps flagged · score 5.0 of 10

Silver labels and silver explanations come from the authors' own T5/Mistral pipeline, with T5 partly trained on Mistral output and XBRL mapping inherited from self-cited work; headline explanation F1 is computed on only 7 of 20 flagged companies.

  1. self definitional [Section 3, Datasets D1, T5 training, and D(SL)_str construction]
    "We used sentence datasets D0, D1 and D(1)_ann to train the T5 model. ... Then we used the MISTRAL-7B-Instruct LLM to label these 1152 sentences in the dataset D_aud as ADV_REMARK or NA. ... If A contains at least 2 sentences both of which are labeled as ADV_REMARK by both MISTRAL and T5 classifier, then we assign the label MISINFO."

    D1 is created by taking the trained BERT model from [3]'s predictions and then labeling the selected 1152 sentences with Mistral; those Mistral labels are part of T5's training data. The silver label 'MISINFO' is then defined as agreement between T5 and Mistral on at least two sentences. The consensus is not between two independent labelers: T5 has been fitted to Mistral's own outputs, so the agreement is partly a model agreeing with its own training signal. Because the same silver labels are used to compute all detection P@20 and explanation F1 values, the benchmark is partially self-confirming rather than an external ground truth.

  2. other [Section 5.6, Tables 11-12]
    "The Table includes only 7 of these companies (out of 20) because for the remaining 13 companies, either (i) the intersection of sentences from audit reports classified by T5 and MISTRAL as ADV REMARK is empty (thus there is no misinformation in these FS as per our criterion); or (ii) the intersection of the generated explanations (by all 5 methods) with silver standard explanation is empty."

    The headline EiForest average F1 of 0.24 in Table 12 is computed on the 7 companies that survived the exclusion, not on the 20 companies originally flagged by SVR M15. The excluded 13 are failures of the same label/explanation pipeline: no T5+Mistral adverse-sentence intersection, or no overlap with the silver explanation. If the 13 are scored as F1=0, the average over the 20 flagged companies falls to about 0.08. The evaluation set is thus selected by the very T5/Mistral mechanism that defines the labels, so the reported explanation performance is not an unbiased estimate for the flagged set.

1 more flagged steps
  1. self citation load bearing [Section 5.5, Constructing the Silver Standard Ground Truth Explanations]
    "The method for assigning 0, 1 or more XBRL categories as the silver standard explanation for explaining why a particular company (i.e., its FS) is labeled MISINFO in the database D(SL)_str is now explained [3]."

    The silver-standard explanation, against which EMD, EiForest, EMI, SHAP and LIME are scored, is produced by the XBRL-category assignment technique inherited from the authors' own prior work [3], with no independent validation reported in this paper. The explanation F1 values therefore measure agreement with a ground truth whose construction is a self-cited method of the same research group, rather than an external benchmark such as restatements or regulatory actions. This is load-bearing because the paper's conclusion that 'explanations identify the variables responsible' rests entirely on this silver standard.

full rationale

The central detection claims are not fully circular by construction: the structured financial statement features used by the AD and regression methods are independent of the audit-report text that generates the silver labels, so a high P@20 is not logically forced by the label definition alone. However, the benchmark quality is compromised by self-generation. T5 is trained in part on Mistral-labeled D1, and the silver 'MISINFO' label is then defined as T5+Mistral agreement, so the two 'independent' labelers are partially the same signal. The explanation evaluation is weaker still: the silver-standard explanations are built from the same T5+Mistral adverse sentences via an XBRL mapping taken from the authors' prior work [3], with no external validation. The reported EiForest F1=0.24 is also computed only on 7 of the 20 flagged companies, excluding 13 pipeline failures; scoring them as zero drops the average to roughly 0.08. These issues make the headline detection and explanation metrics partially self-referential, though the underlying detection task retains enough cross-source independence to avoid a score of 6 or higher.

Assumptions & free parameters 7 free parameters · 5 assumptions · 2 invented entities

The central claim rests on self-generated labels and hand-set thresholds. There are no fitted constants in the mathematical sense, but the silver-label threshold, XBRL matching threshold, and explanation-method parameters are chosen by hand and affect all reported scores. The main domain assumptions are that adverse audit remarks track real misinformation, that misinforming statements are statistical outliers, and that XBRL categories map cleanly onto the 42 extracted variables. No independent gold-standard evidence is provided.

free parameters (7)
  • silver-label agreement threshold = at least 2 adverse sentences agreed by both T5 and Mistral
    Chosen by hand in Section 3 to define the silver-labeled dataset; no sensitivity analysis is provided.
  • XBRL Jaccard similarity threshold theta_1 = 0.6
    Used in Section 5.5 to match adverse sentences to XBRL categories; chosen ad hoc with no reported variation.
  • EMD parameters c and k0 = c=0.5, k0=2
    Set in Section 5.6 for the Mahalanobis-based explanation method; no sensitivity analysis is reported.
  • EiForest parameters T and k = T=1000, top k=5
    Set in Section 5.6 for the isolation-forest explanation method; the paper notes that k affects precision.
  • EMI explanation size L = L=2
    Set in Section 5.6 for the maximal-isolation explanation method; no sensitivity analysis is provided.
  • SHAP and LIME top features = top 5
    Used in Section 5.6 for the baseline explanation methods; the chosen k changes the reported precision.
  • susceptible-variable list for regression = 21 liability-related variables and ratios
    Chosen by hand in Section 4.2 to illustrate regression-based detection; other susceptible variables could be added.
assumptions (5)
  • domain assumption Audit report adverse remarks are reliable indicators that the corresponding financial statement contains misinformation.
    Used to build silver labels in Section 3; no external confirmation from restatements, regulatory actions, or court findings is provided.
  • domain assumption Financial statements containing misinformation appear as statistical anomalies relative to honest peers in the same year.
    This is the core hypothesis behind all anomaly detection methods in Section 4.1.
  • domain assumption XBRL taxonomy categories, promoted to parent categories when needed, correspond to the 42 hand-picked financial variables.
    Required in Section 5.5 to turn adverse sentences into silver-standard explanation ground truth.
  • domain assumption Web-scraped Moneycontrol data for 2,292 companies is complete and correctly parsed.
    All downstream results depend on the scraped variables and audit reports described in Section 3.
  • domain assumption A random forest classifier trained on silver labels is a fair surrogate for generating SHAP and LIME baseline explanations.
    Used in Section 5.6 to create the SHAP and LIME comparison; the classifier inherits all silver-label noise.
invented entities (2)
  • Silver-standard misinformation labels (the labeled dataset)
    purpose: Used as ground truth to compute precision of detection methods and to train baseline classifiers.
    Produced by T5 and Mistral agreement on audit report sentences inside the paper; no confirmed restatements or regulatory actions back the labels.
  • Silver-standard explanation ground truth
    purpose: Used to evaluate EMD, EiForest, EMI, SHAP, and LIME explanations.
    Constructed from the same adverse remarks used to create the silver labels, through XBRL mapping, so it is not an external benchmark.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Financial Audit Assistance using Misinformation Detection and Explanation." pith.science (2026). https://pith.science/paper/4TH2GD4F

@misc{pith2026260717797,
  author       = {Pith},
  title        = {Pith review of: Financial Audit Assistance using Misinformation Detection and Explanation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4TH2GD4F}},
  note         = {Machine review of arXiv:2607.17797}
}
read the original abstract

Financial statements (FS) such as Balance Sheet (BS), Income Statement (IS) and Cash-flow Statement (CS) summarize the annual financial performance of a company. FS are widely used for evaluating corporate governance, credit appraisal, risk analysis, validate taxation, make investment decisions etc. Financial auditing is a complex and knowledge-intensive discipline whose one important aim is ensuring integrity, accuracy, fairness and absence of material misstatement in the published FS. Given the importance of FS, there are incentives to hide, omit or falsify information to misrepresent the true financial health of the company; e.g., reduce tax liabilities, or increase investor confidence. Given the complex, time-consuming and expertise-dependent nature of auditing, auditors would benefit from an AI-assisted system that automatically detects instances of misinformation in the given FS and identify likely sources of this misinformation in the financial data. In this paper, we present unsupervised techniques to identify misinformation in FS, and also generate explanations as to the financial variables that are likely sources of misinformation. The auditor can then explore in more detail the associated data sources and business processes to validate these suggestions. A crucial feature of our approach is the use of past corpus of FS and associated audit reports to generate insights, which help in providing assistance. We demonstrate the efficacy of these techniques on a large corpus of 11,460 FS over 5 years and associated audit reports. This paper integrates and adds more novel contributions over the previously reported research (Shinde et al., 2022)\cite{SVAP22}, (Vaishampayan et al., 2022)\cite{VSPP22}, (Pawar et al., 2023)\cite{PAPV23}, which we have used as the foundation for our AI-assisted Auditor Assistance system.

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 55 canonical work pages

  1. [1]

    In: Proceedings of the International FLAIRS Conference, vol

    Shinde, A., Vaishampayan, S., Apte, M., Palshikar, G.K.: Unsupervised detection of misinformation in financial statements. In: Proceedings of the International FLAIRS Conference, vol. 35 (2022)

    work page 2022

  2. [2]

    In: CIKM-2022 Work- shop on Advances in Interpretable Machine Learning and Artificial Intelligence (AIMLAI 2022) (2022)

    Vaishampayan, S., Shinde, A., Pawde, A., Pawar, S., Apte, M., Palshikar, G.K.: Explainability for misinformation in financial statements. In: CIKM-2022 Work- shop on Advances in Interpretable Machine Learning and Artificial Intelligence (AIMLAI 2022) (2022)

    work page 2022

  3. [3]

    In: Proceedings of the International FLAIRS Conference, vol

    Pawar, S., Apte, M., Pawde, A., Vaishampayan, S., Palshikar, G., Shinde, A.: Dig here! extracting and using knowledge from financial audit reports. In: Proceedings of the International FLAIRS Conference, vol. 36 (2023)

    work page 2023

  4. [4]

    Pearson Publishers, ??? (2016)

    Arens, A., Elder, R., Beasley, M.: Auditing and Assurance Services: An Integrated Approach, 16 edn. Pearson Publishers, ??? (2016)

    work page 2016

  5. [5]

    Wiley, ??? (2007)

    Feldman, M., Libman, A.: Crash Course in Accounting and Financial Statement Analysis, 3rd edn. Wiley, ??? (2007)

    work page 2007

  6. [6]

    Wiley, ??? (2012)

    Drake, P.P., Fabozzi, F.J.: Analysis of Financial Statements, 3rd edn. Wiley, ??? (2012)

    work page 2012

  7. [7]

    Beasley, M., Carcello, J., Hermanson, D.: Fraudulent financial reporting (1987-

    work page 1987

  8. [8]

    Journal of Economic Perspectives 17, 3–26 (2003)

    Healy, P.M., Palepu, K.G.: The fall of enron. Journal of Economic Perspectives 17, 3–26 (2003)

    work page 2003

Show all 58 references
  1. [9]

    Open Journal of Accounting2, 26–38 (2013)

    Bhasin, M.: Corporate accounting fraud: A case study of satyam computers limited. Open Journal of Accounting2, 26–38 (2013)

    2013

  2. [10]

    Wiley, ??? (2020)

    Nigrini, M.J.: Forensic Analytics: Methods and Techniques for Forensic Account- ing Investigations, 2nd edn. Wiley, ??? (2020)

    2020

  3. [11]

    Journal of machine learning research21(140), 1–67 (2020)

    Raffel, C., Shazeer, N., Roberts, A., Lee, K., Narang, S., Matena, M., Zhou, Y., Li, W., Liu, P.J.: Exploring the limits of transfer learning with a unified text-to-text transformer. Journal of machine learning research21(140), 1–67 (2020)

    2020

  4. [12]

    Lennox, C., Li, B.: When are audit firms sued for financial reporting failures and what are the lawsuit outcomes? Contemporary Accounting Research37, 1370–1399 (2019) 27

    2019

  5. [13]

    Journal of Accounting Research48(3), 647–686 (2010)

    Reichelt, K.J., Wang, D.: National and office-specific measures of auditor industry expertise and effects on audit quality. Journal of Accounting Research48(3), 647–686 (2010)

    2010

  6. [14]

    Journal of Accounting, Auditing and Finance30(1), 101–121 (2015)

    Tepalagul, N., Lin, L.: Auditor independence and audit quality: A literature review. Journal of Accounting, Auditing and Finance30(1), 101–121 (2015)

    2015

  7. [15]

    Journal of Accounting and Economics24, 3–37 (1997)

    Basu, S.: The conservatism principle and the asymmetric timeliness of earnings. Journal of Accounting and Economics24, 3–37 (1997)

    1997

  8. [16]

    American Institute of Certified Public Accountants, ??? (1999)

    Beasley, M.S., Carcello, J.V., Hermanson, D.R., Treadway Commission, C.,et al.: Fraudulent Financial Reporting: 1987-1997: an Analysis of US Public Companies. American Institute of Certified Public Accountants, ??? (1999)

    1999

  9. [17]

    Managerial Auditing Journal17, 179–191 (2002)

    Spathis, C.T.: Detecting false financial statements using published data: some evidence from greece. Managerial Auditing Journal17, 179–191 (2002)

    2002

  10. [18]

    International journal of computa- tional intelligence3, 104–110 (2006)

    Kotsiantis, S., Koumanakos, E., Tzelepis, D., Tampakas, V.: Forecasting fraud- ulent financial statements using data mining. International journal of computa- tional intelligence3, 104–110 (2006)

    2006

  11. [19]

    SpringerPlus5, 1–16 (2016)

    Chen, S.: Detection of fraudulent financial statements using the hybrid data mining approach. SpringerPlus5, 1–16 (2016)

    2016

  12. [20]

    Sustainability10, 513 (2018)

    Jan, C.-l.: An effective financial statements fraud detection model for the sustain- able development of financial markets: Evidence from taiwan. Sustainability10, 513 (2018)

    2018

  13. [21]

    In: IEEE Symposium Series on Computational Intelligence (SSCI), pp

    Mongwe, W.T., Malan, K.M.: The efficacy of financial ratios for fraud detec- tion using self organising maps. In: IEEE Symposium Series on Computational Intelligence (SSCI), pp. 1100–1106 (2020)

    2020

  14. [22]

    Asian Journal of Accounting Research4(2019)

    Lokanan, M., Tran, V., Vuong, N.H.: Detecting anomalies in financial statements using machine learning algorithm: The case of vietnamese listed firms. Asian Journal of Accounting Research4(2019)

    2019

  15. [23]

    In: Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp

    Li, H., Yang, Q., Cao, Y., Yao, J., Luo, P.: Cracking tabular presentation diversity for automatic crosschecking over numerical facts. In: Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 2599–2607 (2020)

    2020

  16. [24]

    In: 24th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD2018), pp

    Roux, D., Perez, B., Moreno, A., Pilar Villamil, M., , Figueroa, C.: Tax fraud detection for underreporting declarations using an unsupervised machine learning approach. In: 24th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD2018), pp. 215–222 (2018)

    2018

  17. [25]

    In: 19th International Database Engineering and Applications Symposium (IDEAS 2015), pp

    Matos, T., Macedo, J.A.F., Monteiro, J.M.: An empirical method for discovering 28 tax fraudsters: A real case study of brazilian fiscal evasion. In: 19th International Database Engineering and Applications Symposium (IDEAS 2015), pp. 41–48 (2015)

    2015

  18. [26]

    Expert Systems with Applications 40, 1427–1436 (2013)

    Gonzalez, P., Velasquez, J.: Characterization and detection of taxpayers with false invoices using data mining techniques. Expert Systems with Applications 40, 1427–1436 (2013)

    2013

  19. [27]

    Expert Systems with Applications 39, 8769–8777 (2012)

    Wu, R.-S., Ou, C., Lin, H.-Y., Chang, S.-I., Yen, D.: Using data mining technique to enhance tax evasion detection performance. Expert Systems with Applications 39, 8769–8777 (2012)

    2012

  20. [28]

    In: VLDB, pp

    Zhang, J., Lou, M., Ling, T.W., Wang, H.: HOS-miner: A system for detecting outlying subspaces of high-dimensional data. In: VLDB, pp. 1265–1268 (2004)

    2004

  21. [29]

    arXiv preprint arXiv:2005.02637 (2020)

    Samariya, D., Ma, J., Aryal, S.: A comprehensive survey on outlying aspect mining methods. arXiv preprint arXiv:2005.02637 (2020)

    2020 arXiv

  22. [30]

    Data Mining and Knowledge Discovery29(5), 1116– 1151 (2015)

    Duan, L., Tang, G., Pei, J., Bailey, J., Campbell, A., Tang, C.: Mining outlying aspects on numeric data. Data Mining and Knowledge Discovery29(5), 1116– 1151 (2015)

    2015

  23. [31]

    Data Mining and Knowledge Discovery30(6), 1520–1555 (2016)

    Vinh, N.X., Chan, J., Romano, S., Bailey, J., Leckie, C., Ramamohanarao, K., Pei, J.: Discovering outlying aspects in large datasets. Data Mining and Knowledge Discovery30(6), 1520–1555 (2016)

    2016

  24. [32]

    In: PAKDD, pp

    Vinh, N.X., Chan, J., Bailey, J., Leckie, C., Ramamohanarao, K., Pei, J.: Scalable outlying-inlying aspects discovery via feature ranking. In: PAKDD, pp. 422–434 (2015). Springer

    2015

  25. [33]

    In: International Conference on Data Engineering, pp

    Dang, X.H., Assent, I., Ng, R.T., Zimek, A., Schubert, E.: Discriminative features for identifying and interpreting outliers. In: International Conference on Data Engineering, pp. 88–99 (2014). IEEE

    2014

  26. [34]

    Data Mining and Knowledge Discovery31(1), 134–163 (2017)

    Angiulli, F., Fassetti, F., Manco, G., Palopoli, L.: Outlying property detection with numerical attributes. Data Mining and Knowledge Discovery31(1), 134–163 (2017)

    2017

  27. [35]

    In: International Confer- ence on Big Data, pp

    Zhang, X., Marwah, M., Lee, I.-t., Arlitt, M., Goldwasser, D.: Ace–an anomaly contribution explainer for cyber-security applications. In: International Confer- ence on Big Data, pp. 1991–2000 (2019). IEEE

    2019

  28. [36]

    ACM Transactions on Knowledge Discovery in Databases13(1), 1–22 (2019)

    Siddiqui, M.A., Fern, A., Dietterich, T.G., Wong, W.-K.: Sequential feature expla- nations for anomaly detection. ACM Transactions on Knowledge Discovery in Databases13(1), 1–22 (2019)

    2019

  29. [37]

    29 Expert Systems with Applications149, 113187 (2020)

    Kopp, M., Pevn` y, T., Holeˇ na, M.: Anomaly explanation with random forests. 29 Expert Systems with Applications149, 113187 (2020)

    2020

  30. [38]

    In: ECML-PKDD, pp

    Gupta, N., Eswaran, D., Shah, N., Akoglu, L., Faloutsos, C.: Beyond outlier detection: Lookout for pictorial explanation. In: ECML-PKDD, pp. 122–138 (2018)

    2018

  31. [39]

    Transactions on Neural Networks and Learning Systems28(11), 2660–2673 (2016)

    Samek, W., Binder, A., Montavon, G., Lapuschkin, S., M¨ uller, K.-R.: Evaluating the visualization of what a deep neural network has learned. Transactions on Neural Networks and Learning Systems28(11), 2660–2673 (2016)

    2016

  32. [40]

    In: International Conference on Pattern Recognition and Machine Intelligence, pp

    Yadav, A., Jha, C., Sharan, A., Vaish, V.: Sentiment analysis of financial news using unsupervised and supervised approach. In: International Conference on Pattern Recognition and Machine Intelligence, pp. 311–319 (2019). Springer

    2019

  33. [41]

    In: M´ etais, E., Meziane, F., Saraee, M., Sugumaran, V., Vadera, S

    Patil, S., Palshikar, G.K.: Surveycoder: A system for classification of survey responses. In: M´ etais, E., Meziane, F., Saraee, M., Sugumaran, V., Vadera, S. (eds.) Proc. of Natural Language Processing and Information Systems, pp. 417–420. Springer, ??? (2013)

    2013

  34. [42]

    In: International Conference on Intelligent Text Processing and Computational Linguistics, pp

    Patil, S., Ravindran, B.: Active learning based weak supervision for textual survey response classification. In: International Conference on Intelligent Text Processing and Computational Linguistics, pp. 309–320 (2015)

    2015

  35. [43]

    In: 2022 IEEE/ACM 1st International Workshop on Natural Language-Based Software Engineering (NLBSE), pp

    Devine, P., Blincoe, K.: Unsupervised extreme multi label classification of stack overflow posts. In: 2022 IEEE/ACM 1st International Workshop on Natural Language-Based Software Engineering (NLBSE), pp. 1–8 (2022)

    2022

  36. [44]

    arXiv preprint arXiv:2203.06482 (2022)

    Loukas, L., Fergadiotis, M., Chalkidis, I., Spyropoulou, E., Malakasiotis, P., Androutsopoulos, I., Paliouras, G.: FiNER: Financial numeric entity recognition for XBRL tagging. arXiv preprint arXiv:2203.06482 (2022)

    2022 arXiv

  37. [45]

    In: Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing

    Reimers, N., Gurevych, I.: Sentence-bert: Sentence embeddings using siamese bert-networks. In: Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, ??? (2019).https://arxiv.org/abs/1908.10084

    2019 arXiv

  38. [46]

    Journal of Machine Learning Research20, 1–7 (2019)

    Zhao, Y., Nasrullah, Z., Li, Z.: PyOD: A python toolbox for scalable outlier detection. Journal of Machine Learning Research20, 1–7 (2019)

    2019

  39. [47]

    In: 6th Pacific-Asia Conf

    Tang, J., Chen, Z., Fu, A.W.-c., Cheung, D.: A robust outlier detection scheme for large data sets. In: 6th Pacific-Asia Conf. on Knowledge Discovery and Data Mining (PAKDD 2001), pp. 6–8 (2001)

    2001

  40. [48]

    ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM)15(4), 1–23 (2019) 30

    Zhong, S.-H., Wang, Y., Ren, T., Zheng, M., Liu, Y., Wu, G.: Steganographer detection via multi-scale embedding probability estimation. ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM)15(4), 1–23 (2019) 30

    2019

  41. [49]

    KI-2012: poster and demo track1, 59–63 (2012)

    Goldstein, M., Dengel, A.: Histogram-based outlier score (hbos): A fast unsu- pervised anomaly detection algorithm. KI-2012: poster and demo track1, 59–63 (2012)

    2012

  42. [50]

    Accessed on May 17, 2025 (2025)

    scikit-learn: Random Forest Regression. Accessed on May 17, 2025 (2025). https://scikit-learn.org/stable/modules/generated/sklearn.ensemble. RandomForestRegressor.html

    2025

  43. [51]

    Data & Knowledge Engineering137, 101946 (2022)

    Tchaghe, V.Y., Smits, G., Pivert, O.: Anomaly explanation: A review. Data & Knowledge Engineering137, 101946 (2022)

    2022

  44. [52]

    Data Science and Engineering7, 120–135 (2022)

    Samariya, D., Ma, J.a.: A new dimensionality-unbiased score for efficient and effective outlying aspect mining. Data Science and Engineering7, 120–135 (2022)

    2022

  45. [53]

    In: 27th Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD2023) - Part I, pp

    L¨ udtke, S., Bartelt, C., Stuckenschmidt, H.: Outlying aspect mining via sum- product networks. In: 27th Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD2023) - Part I, pp. 27–38 (2023)

    2023

  46. [54]

    Journal of the Royal Statistical Society: Series B30, 205–247 (1968)

    Dempster, A.P.: A generalization of bayesian inference. Journal of the Royal Statistical Society: Series B30, 205–247 (1968)

    1968

  47. [55]

    In: 8th ICDM, pp

    Liu, F.T., Ting, K.M., Zhou, Z.-H.: Isolation forest. In: 8th ICDM, pp. 413–422 (2008). IEEE

    2008

  48. [56]

    Kuhn, H.W., Tucker, A.W.: Contributions to the Theory of Games vol. 28. Princeton University Press, ??? (1953)

    1953

  49. [57]

    In: Proceedings of the 22nd ACM SIGKDD Inter- national Conference on Knowledge Discovery and Data Mining, pp

    Ribeiro, M.T., Singh, S., Guestrin, C.: ” why should i trust you?” explaining the predictions of any classifier. In: Proceedings of the 22nd ACM SIGKDD Inter- national Conference on Knowledge Discovery and Data Mining, pp. 1135–1144 (2016) 31

    2016

  50. [1997]

    Technical report, American Institute of Certified Public Accountants (1999)

    an analysis of US public companies. Technical report, American Institute of Certified Public Accountants (1999)

    1999

Pith tools

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