Presentation given by Rommel Carvalho at the 5th Uncertainty Reasoning for the Semantic Web Workshop at the 8th International Semantic Web Conference in 2009.
Paper: Probabilistic Ontology and Knowledge Fusion for Procurement Fraud Detection in Brazil
Abstract: To cope with society’s demand for transparency and corruption prevention, the Brazilian Federal General Comptroller (CGU) has carried out a number of actions, including: awareness campaigns aimed at the private sector; campaigns to educate the public; research initiatives; and regular inspections and audits of municipalities and states. Although CGU has collected information from hundreds of different sources - Revenue Agency, Federal Police, and others - the process of fusing all this data has not been efficient enough to meet the needs of CGU’s decision makers. Therefore, it is natural to change the focus from data fusion to knowledge fusion. As a consequence, traditional syntactic methods must be augmented with techniques that represent and reason with the semantics of databases. However, commonly used approaches fail to deal with uncertainty, a dominant characteristic in corruption prevention. This paper presents the use of Probabilistic OWL (PR-OWL) to design and test a model that performs information fusion to detect possible frauds in procurements involving Federal money. To design this model, a recently developed tool for creating PR-OWL ontologies was used with support from PR-OWL specialists and careful guidance from a fraud detection specialist from CGU.
Measures of Central Tendency: Mean, Median and Mode
URSW 2009 - Probabilistic Ontology and Knowledge Fusion for Procurement Fraud Detection in Brazil
1. Probabilistic Ontology and Knowledge
Fusion for Procurement Fraud Detection
in Brazil
Rommel Carvalho, Kathryn Laskey, and Paulo Costa
George Mason University
Marcelo Ladeira, Laécio Santos, and Shou Matsumoto
Universidade de Brasília
Paper - Uncertainty Reasoning for the Semantic Web
URSW - ISWC
9. Introduction
Brazilian Office of the Comptroller General
(CGU) primary mission
Prevent and detect irregularities (corruption)
Gather information from a variety of sources
Combine the information
Then evaluate whether further action is necessary
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 4
10. Introduction
Brazilian Office of the Comptroller General
(CGU) primary mission
Prevent and detect irregularities (corruption)
Gather information from a variety of sources
Combine the information
Then evaluate whether further action is necessary
Problem
Information explosion
Growing Acceleration Program (PAC)
250 billion dollars - 1,000+ projects only in SP
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 4
12. Introduction
MEBN
Represent and reason with uncertainty about any
propositions that can be expressed in first-order logic
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 5
13. Introduction
MEBN
Represent and reason with uncertainty about any
propositions that can be expressed in first-order logic
PR-OWL
Uses MEBN logic to provide a framework for building
probabilistic ontologies
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 5
14. Introduction
MEBN
Represent and reason with uncertainty about any
propositions that can be expressed in first-order logic
PR-OWL
Uses MEBN logic to provide a framework for building
probabilistic ontologies
Fraud Detection and Prevention Model
Uses MEBN and PR-OWL
Proof of concept
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 5
28. Procurement Fraud Detection
A major source of corruption is the procurement
process
Laws attempt to ensure a competitive and fair process
Perpetrators find ways to turn the process to their
advantage while appearing to be legitimate
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 11
29. Procurement Fraud Detection
A major source of corruption is the procurement
process
Laws attempt to ensure a competitive and fair process
Perpetrators find ways to turn the process to their
advantage while appearing to be legitimate
Specialist from CGU (Mário Spinelli)
Structured the different kinds of procurement frauds found
in the past years
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 11
31. Procurement Fraud Detection
Types of fraud
Characterized by criteria
Principle of competition is violated when we have
Owners who work as a front (usually someone with little or no education)
Use of accounting indices that are not common
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 12
32. Procurement Fraud Detection
Types of fraud
Characterized by criteria
Principle of competition is violated when we have
Owners who work as a front (usually someone with little or no education)
Use of accounting indices that are not common
Ultimate goal
Structure the specialist knowledge in a way that an
automated system can reason with the evidence in a manner
similar to the specialist
Support current specialists
Train new ones
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 12
34. Procurement Fraud Detection
Realistic goal (this paper)
Proof of concept
Selected just a few criteria
Why Semantic Web?
Propose an overall architecture for collecting data, reasoning
with uncertainty (model designed), and reporting alerts
Ask specialists to analyze results (subjective)
No massive data used
Show that new criteria can be easily incorporated
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 13
35. Why Semantic Web?
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion
14
* AAA, open world, and nonunique naming - RIS environment
36. Why Semantic Web?
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion
14
* AAA, open world, and nonunique naming - RIS environment
37. Why Semantic Web?
audits and inspections
(Procurement)
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion
14
* AAA, open world, and nonunique naming - RIS environment
38. Why Semantic Web?
audits and inspections
(Procurement) socio-economic
(Person)
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion
14
* AAA, open world, and nonunique naming - RIS environment
39. Why Semantic Web?
audits and inspections
(Procurement) socio-economic
(Person)
criminal history
(Person)
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion
14
* AAA, open world, and nonunique naming - RIS environment
40. Why Semantic Web?
audits and inspections
(Procurement) socio-economic
(Person)
? criminal history
(Person) (Person)
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion
14
* AAA, open world, and nonunique naming - RIS environment
43. Architecture
Informaion Gathering
Public Notices - Data
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 15
44. Architecture
Informaion Gathering
DB - Information
Public Notices - Data
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 15
45. Architecture
Informaion Gathering
DB - Information
Public Notices - Data Design - UnBBayes
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 15
46. Architecture
Informaion Gathering
DB - Information
Public Notices - Data Design - UnBBayes
Inference - Knowledge
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 15
47. Architecture
Informaion Gathering
DB - Information
Public Notices - Data Design - UnBBayes
Report for Decision Makers Inference - Knowledge
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 15
48. Architecture
Informaion Gathering
DB - Information
Public Notices - Data Design - UnBBayes
Report for Decision Makers Inference - Knowledge
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 15
54. Results
Non suspect procurement:
0.01% that the procurement was directed to a specific company by using accounting
indices;
0.10% that the procurement was directed to a specific company.
Suspect procurement:
55.00% that the procurement was directed to a specific company by using accounting
indices;
29.77%, when the information about demanding experience in only one contract was
omitted, and 72.00%, when it was given, that the procurement was directed to a specific
company.
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 20
59. Conclusion
Correct conclusion for both suspicious and non-
suspicious cases
Results are encouraging
Suggesting that a fuller development of our proof of concept is promising
Needs more testing, especially with real data for validating the conclusions
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 24
60. Conclusion
Correct conclusion for both suspicious and non-
suspicious cases
Results are encouraging
Suggesting that a fuller development of our proof of concept is promising
Needs more testing, especially with real data for validating the conclusions
Advantages
Impartiality in the judgment
Scalability
Joint analysis of large volumes of indicators
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 24
62. Conclusion
Future work
Choose/add new criteria
Collect more data for validation of the model
Will probably required fusion of data from different
agencies
Good for assessing the usefulness of ontologies and the SW
Introduction - MEBN and PR-OWL - Procurement Fraud Detection - Conclusion 25