Design of Exogenous Neuro-Structure for Dynamical Analysis of Nonlinear Fractional Financial Crime Systems.

The contributions of AI-based applications in monitoring real-time financial transactions, and detecting fraudulent activity by scrutinizing consumer behavior, transaction patterns, and other relevant measures are worth mentioning for potential threats identification in the fractional financial crime population dynamics. Leveraging these financial crime systems in terms of population dynamics with the exploitation of supervised Nonlinear Autoregressive Exogenous Networks Optimized with the Bayesian Regularization (NARX-BR) procedures for attaining sufficient accuracy and flexibility for the approximate solutions of a fractional variant of stiff Nonlinear Financial Crime Population Dynamics (NFCPDs) differential system. The population dynamics for the financial crime model are classified mainly into susceptible persons, financial criminals, individuals being prosecuted individuals under prosecution, imprisoned persons, and honest individuals by law. The acquisition of synthetic data generated with Grünwald–Letnikov (GL) fractional operator for the multi-layer structure execution of NARX-BR procedure for solving NFCPDs for varying financial crime parameters, such as influence rate, recruitment rate, conversion rate to honest people, freedom rate, financial criminal prosecution rate per capita, percentage of discharge rate from prosecution, transition rate to prison, discharge and acquittal rate from prosecutions. The estimated outcomes of NARX-BR and the calculated numerical solutions of NFCPDs consistently overlap implying that the error between the results is approximately equal to zero. The effectiveness of model performance is assessed through a variety of evaluation metrics, that include minimization of mean square error-based objective function, adaptive regulating parameters of the optimization algorithm, error distribution plots, regression studies, error endogeneity, and cross-correlation analyses. This study contributes to integrating fractional calculus with the knacks of innovative AI and open paths to provide data-driven efficient solution-based policy recommendations in the field of financial crime population dynamics. [ABSTRACT FROM AUTHOR]