Advanced Analytics: 11 Mathematical Fraud Detection Methods

11 mathematical approaches across 8 domains applied to federal fraud data: PageRank and network topology (graph theory), Shannon entropy (information theory), Bayesian risk scoring, Markov chains, SVD latent factor analysis, Benford's Law, linear programming for investigation portfolio optimization, and Nash equilibrium modeling (game theory).

The optimal investigation portfolio selects 5 cases from 29 candidates using 1,966 of 2,000 budgeted hours, with expected recovery of $6.96 billion and potential qui tam rewards of $1.39 billion.