Varieties Statistical Fraud Models 30 Models in 30

Varieties Statistical Fraud Models: 30 Models in 30 Minutes Daniel Finnegan, CFE ISO Innovative Analytics Quality Planning Corporation

Benford’s Law in Accounting Fraud

Tests for Manufacture Numbers u u u u Frequency or equidistribution test (possible elements should occur with equal frequency); Serial test (pairs of elements should be equally likely to be in descending and ascending order); Gap test (runs of elements all greater or less than some fixed value should have lengths that follow a binomial distribution); Coupon collector's test (runs before complete sets of values are found should have lengths that follow a definite distribution); Permutation test (in blocks of elements possible orderings of values should occur equally often); Runs up test (runs of monotonically increasing elements should have lengths that follow a definite distribution); Maximum-of-t test (maximum values in blocks of elements should follow a power-law distribution).

IRS Audit Selection System 1964 Rule-Based Scoring System 1970’s TCMP Statistical Audit System 2003 NRP System: A. Random Audits of Sample of Returns B. Identification of Returns “In Need of Examine” C. Statistical Model of DIF score of “Probability of Need to Examine” D. Monitoring and Update of System

Text Mining for Fraudulent Medical Bills Search for identical typos u Search for identical prognosis u Search for date discrepancies u § Holidays § Claimant out of town/dead

Medical Usage Pattern Fraud Analysis Uniformly high numbers of treatments (Normed on Diagnosis) u High number of modalities per treatment u Few Patients Recover Quickly u Low Percentage of Objective Injuries u Treatment Ends Abruptly at Payment of Claim u

FAIS Money Laundering Statistical Detection Link Analysis with Known Criminal Elements u Pattern Analysis such as Large Sum Deposited and Immediately Withdrawn u Benford Distribution of Deposits and Withdrawals u Circular Movements of Funds u

Network Analysis of Auto Accidents

Staged Accident Ring

Sequential Handling of Questionable Claims u u Random Sample of 3, 000 BI Claims Decision Flow Model

Timing Claims Curves

Other Threshold Fraud Models Adding Coverage for Comp u Two-Year New Vehicle Replacement u School Lunch Eligibility u

Deviant Purchase Patterns for Credit Card Fraud u u Identification of Individual Purchase Patterns (Neural Net Models) Identification of Typical Fraud Purchase Patterns (Electronics, International Spending) Movement out of Typical Toward Fraud Patterns Expert Patterns Such Geographic Dispersion of Purchases

Geographic Analysis of Staged Accidents

Geographic Analysis of Staged Accidents

Driver’s License Translator Fraud u Pass Rate: § 51% vs 95+% u Time to Complete § 30 -60 Minutes vs 10 -15 Minutes

Insider Stock Dealing u u u Mon. ITARS: Fuzzy Logic, Neural Nets, Genetic Algorithms for London Stock Exchange Advanced Detection System (ADS) for Nasdaq matches rule-based sequential trading patterns SONAR matches wire stories to stock trading using pattern analysis to detect stock manipulation

WC Premium Audit Selection Model u u u Statistical Modeling of 4 Years of Audit Results Holdback of 5 th Year of Results Combined Expert Theory and Inductive Modeling Final Model Built with Multiple Statistical Methods: § Decision Trees, MARS, GLM Model Concentrated on Key Ratios by Industry Results more than Doubled Audit Returns

University Student Aid Fraud Very High and Similar Hardship Deductions (High Medical Bills) u Identical Applications for Student Financial Aid (High Aid with No Audit) u Fraud Clusters by Successful Sports Teams u

Work Load Analysis of Medical Billing Fraud Psychiatrist billing 80 hour work days u Billing on 365 day years u Billing from distant locations u Billing for 200 patients per day u

Adjuster – Vendor Pairing Models Billing Pattern Analysis for 5 Million Claims and 12 Million Payments u Dozen Questionable Patterns Identified: u § Relative High Payment Average for Adjuster and Vendor § Identification of Vendors with Multiple Payments to PO Box with Single Adjuster

Social Security Disability Model Random Sample File Review u Identified Decision Errors/Fraud u Built Multiple Models u § § u Econometric Decision Trees, GLM, Hybrid Rule Violation Decision Maker Focused Final Artificial Intelligence Model

Sales Agent Rating Models u Sales Agents Mileage Model § Low to Expectations § Below Rating Cut Points u Missing Drivers § Teenagers Low to Expectations § High Permissive Use Claims u Frequent Claims After Comp Added

Food Stamp Store Investigation System Prior System Viewed as a Success u Random Investigation of 2, 000 Stores u Statistical Analysis of Discovered Violations u

Food Stamp Investigation Outcomes

VIPER System Statistical Pattern Targeting u Random Component for Updating u Geographic Clustering Component u Tripled Discovered Violations u Doubled Investigator Productivity u

Thresholding Cell Phone Accounts u u u 6 -8 Percent Cell Phone Costs Fraudulent High Volume of Calls and Turnover of Fraud Requires Rapid Response Account “Thresholding” Process Used § 30 -Day, Fraud Free, Norming Process § Account Specific Expert Rules on Duration, Location, Timing § Calls Scored Statistical Distance from Norms § Percent of Potential Fraud Calls Monitored § Norms Constantly Updated

Identity Theft Scoring System Includes Variety of Data Matching and Pattern Analysis Variables u High Numbers of Credit Card or Cell Phone Applications from Address u Identity Variable Conflicts u Mail Drop Address u Impossible SSN Dead, Issued Before Born, Un-issued, Impossible

Statistical Adjuster Assignment Models Review of Areas of Fraud Loss u Identification of Best Practices for Handling Questionable Claims u Sample Investigation of Matched Samples of 1, 500 Standard Handling and 1, 500 Enhanced Handling u Statistical Modeling of Handling Gains u

Statistical Adjuster Assignment Models

Common Elements of Successful Statistical Fraud Control Statistical Methods Selected to Fit the Problem (One Size Does Not Fit All) u High Input from Substance Area Experts u Feedback Loop Evaluates and Updates System u Strong Integration with Operations u