Chp1 Fraud:Detection,Prevention and analytics
1.1 INTRODUCTION
Fraud is an uncommon, well-considered, imperceptibly concealed, time-evolving and often carefully organized crime which appears in many types of forms.
Unexhaustive list of fraud categories and types:
Credit Card fraud
1. Application fraud:obtaining new cards from issuing companies by using false personal information, and then spending as much as possible in a short space of time.
2. Behavioral fraud:
Insurance fraud
Corruption
Counterfeit
Product warranty fraud
Healthcare fraud
Telecommunications fraud
Money laundering
Click fraud
Identity theft
Tax evasion
Plagiarism
unravel
recompense
loophole
chameleon
camouflage
stick insects
anomaly
subtlety
premise
ingredient
prominent
overlap
marginal
- Fraud Triangle
- Pressure
- Opportunity
- Rationalization
Fraud is a social phenomenon in the sense that the potential benefits for the fraudsters come at the expense of the victims.
Estimated size and the financial impact of fraud:
5% of one’s revenues to fraud each year for A typical organization.
40 billion/year for the total cost of insurance fraud in USA.
73 billion/year is costing for the National Fraud Authority
7 cents per 100 dollars of transaction due to fraud for Credit Card Company.
The average size of the informal economy, as a percent of official GNI in the year 2000, 41% in developing countries, 38% in transition countries, 18% in OECD countries
1.2 FRAUD DETECTION AND PREVENTION
Detection - ex post approach
Prevention - ex ante approach
1.3 BIG DATA FOR FRAUD DETECTION
corrective measures
preventive measures
fraud becomes easier to detect ,because more similiar fraud cases will occur in time, and better detection techniques are being developed, and growing data availability
1.4 DATA-DRIVEN FRAUD DETECTION
A shift is taking place toward data-driven or statistically based fraud-detection methodologies from classic,expert-based fraud-detection approaches.
Reasons: Precision/Operational efficiency/Cost efficiency
1.5 FRAUD-DETECTION TECHNIQUES
Unsupervised learning techniques: descriptive anaysis(Outlier detection techniques):finding behaviors that deviates from normal behavior(detecting anomalies)--to recognize novel fraud pattern
Supervised learning techniques: predictive analytics:aim to learn from historical information or observations in order to retriveve patterns that allow differentiating between normal and fraudulent behavior. -- need historical data and labels
Social network analysis:learning and detecting characteristics of fraudulent behavior in a network of linked entities.
To develop a fraud-detection system:
step1. an expert-based rule engine may be developed
step2. descriptive analytics may be complemented
step3. predictive and social network analytics