A Statistical Method for Corrupt Agents Detection

Abstract

The statistical method is used to identify the hidden leaders of the corruption structure. The method is based on principal component analysis (PCA), linear regression, and Shannon information. It is applied to study the time series data of corrupt financial activity. Shannon's quantity of information is specified as a function of two arguments: a vector of hidden corruption factors and a subset of corrupt agents. Several optimization problems are solved to determine the contribution of corresponding corrupt agents to the total illegal behavior. An illustrative example is given. A convenient algorithm for computing the covariance matrix with missing data is proposed.