High Quality True-Positive Prediction for Fiscal Fraud Detection

Basta, Stefano; Fassetti, Fabio; Guarascio, Massimo; Manco, Giuseppe; Giannotti, Fosca; Pedreschi, Dino; Spinsanti, Laura; Papi, Gianfilippo; Pisani, Stefano

doi:10.1109/icdmw.2009.59

Cited by 10 publications

(7 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The authors examine C5.0 algorithm for learning decision tress to build a classification model for detecting tax evasion in Italy. Other supervised machine learning approaches for detecting fraudulent tax behavior investigated in the literature include random forests (Mittal et al, 2018), rule-based classification (Basta et al, 2009) and Bayesian networks (da Silva et al, 2016).…”

Section: Related Work and Our Contributionsmentioning

confidence: 99%

“…The application of unsupervised learning is mostly relying on various anomaly detection algorithms. These include various supervised, semi-supervised and unsupervised machine learning methods, including decision trees, spectral clustering methods, neural networks, graph-based methods, etc, for example see Wu et al (2012), Matos et al (2015), Bonchi et al (1999), (Basta et al, 2009), (da Silva et al, 2016), Castellón González and Velásquez (2013), Tian et al (2016), de Roux et al (2018), Mehta et al (2020). However, most existing anomaly detection studies focus on devising accurate detection models only, ignoring the capability of providing explanation of the identified anoma-lies (Pang et al, 2020).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Tax Evasion Risk Management Using a Hybrid Unsupervised Outlier Detection Method

Savić¹,

Atanasijević²,

Jakovetić³

et al. 2021

Preprint

View full text Add to dashboard Cite

Big data methods are becoming an important tool for tax fraud detection around the world. Unsupervised learning approach is the dominant framework due to the lack of label and ground truth in corresponding data sets although these methods suffer from low interpretability. HUNOD, a novel hybrid unsupervised outlier detection method for tax evasion risk management, is presented in this paper. In contrast to previous methods proposed in the literature, the HUNOD method combines two outlier detection approaches based on two different machine learning designs (i.e, clustering and representational learning) to detect and internally validate outliers in a given tax dataset. The HUNOD method allows its users to incorporate relevant domain knowledge into both constituent outlier detection approaches in order to detect outliers relevant for a given economic context. The interpretability of obtained outliers is achieved by training explainable-by-design surrogate models over results of unsupervised outlier detection methods. The experimental evaluation of the HUNOD method is conducted on two datasets derived from the database on individual personal income tax declarations collected by the Tax Administration of Serbia. The obtained results show that the method indicates between 90% and 98% internally validated outliers depending on the clustering configuration and employed regularization mechanisms for representational learning.

show abstract

Section: Related Work and Our Contributionsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Tax Evasion Risk Management Using a Hybrid Unsupervised Outlier Detection Method

Savić¹,

Atanasijević²,

Jakovetić³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…An association rule technique to identify VAT evasion tax reports is applied in [33]. Other rule-based approaches are proposed for detecting fraudulent VAT credit claims [34] and for the identification of frequent fraud patterns in the Brasilian fiscal environment [35]. Some tax administrations use clustering techniques to detect groups of taxpayers characterized by non compliance behavior [36], [37].…”

Section: B Data Mining and Machine Learning Approachesmentioning

confidence: 99%

Combining Network Visualization and Data Mining for Tax Risk Assessment

et al. 2020

View full text Add to dashboard Cite

This paper presents a novel approach, called MALDIVE, to support tax administrations in the tax risk assessment for discovering tax evasion and tax avoidance. MALDIVE relies on a network model describing several kinds of relationships among taxpayers. Our approach suitably combines various data mining and visual analytics methods to support public officers in identifying risky taxpayers. MALDIVE consists of a 4-step pipeline: (i) A social network is built from the taxpayers data and several features of this network are extracted by computing both classical social network indexes and domain-specific indexes; (ii) an initial set of risky taxpayers is identified by applying machine learning algorithms; (iii) the set of risky taxpayers is possibly enlarged by means of an information diffusion strategy and the output is shown to the analyst through a network visualization system; (iv) a visual inspection of the network is performed by the analyst in order to validate and refine the set of risky taxpayers. We discuss the effectiveness of the MALDIVE approach through both quantitative analyses and case studies performed on real data in collaboration with the Italian Revenue Agency. INDEX TERMS Tax risk assessment, tax evasion discovery, network visualization, data mining, human-computer interaction.

show abstract

“…De ellas en un 42,81% hubo conformidad del interesado y en un 34,76% se instruyó un expediente sancionador. En Basta et al (2009), en un estudio transversal de las administraciones tributarias se ofrece una cifra similar, el 85,29%. Dado que el control de los tributos exige muchos recursos, el porcentaje de los contribuyentes controlados es bajo.…”

Section: El Problema Tributariounclassified

“…La mayoría de las administraciones tributarias avanzadas han publicado los resultados de proyectos significativos realizados orientado al control del fraude: control del fraude en IVA (Vanhoeyveld et al, 2020), uso indebido de créditos tributarios en Italia (Agenzia delle Entrate, 2007;Andini et al,2018;Basta et al, 2009), uso de minería de datos y redes neuronales para el control tributario en Finlandia (Titan, 2012), el uso de árboles de clasificación en EE.UU (Murthy, 1988), control del fraude a la Seguridad Social en Bélgica (Baesen et al, 2015), y el control de los impuestos directos en Brasil (Matos et al, 2016).…”

Section: Uso De ML En El Control Del Fraude Tributariounclassified

Análisis de fraude en multigrafos : el caso del fraude tributario

García¹

View full text Add to dashboard Cite

show abstract

High Quality True-Positive Prediction for Fiscal Fraud Detection

Cited by 10 publications

References 12 publications

Tax Evasion Risk Management Using a Hybrid Unsupervised Outlier Detection Method

Tax Evasion Risk Management Using a Hybrid Unsupervised Outlier Detection Method

Combining Network Visualization and Data Mining for Tax Risk Assessment

Análisis de fraude en multigrafos : el caso del fraude tributario

Contact Info

Product

Resources

About