Application of fuzzy bayesian logical-probability modeling algorithms in creating cross-platform financial monitoring software

DOI: 10.31673/2412-9070.2026.017413

Abstract

A decision support model in the field of financial monitoring is proposed, based on a neural-network implementation of the Bayesian logical-probabilistic model of fuzzy inference. The approach employs Z-numbers whose components represent fuzzy extensions of probability distributions, enabling simultaneous consideration of two key types of uncertainty – fuzziness and randomness – together with the reliability of incoming information. This makes the model suitable for complex environments in which data are incomplete, inconsistent or derived from heterogeneous sources. Such conditions are typical for financial monitoring systems, where operational decisions must be made on the basis of dynamically changing data flows and varying levels of confidence in the information received. The proposed model integrates elements of fuzzy logic, probabilistic reasoning and neural network approximation, which allows the system to adaptively learn decision patterns and improve the accuracy of assessments under mixed uncertainty. The neural-network component enhances the model’s ability to generalize and recognize latent structures in transactional behavior, while the Bayesian framework maintains interpretability and supports transparent reasoning about risks. An illustrative example demonstrates the operation of the model for detecting atypical financial behavior, confirming its effectiveness in supporting expert decision-making within dynamic monitoring tasks and in situations where traditional deterministic rules fail to capture subtle deviations. Conceptual principles for constructing a flexible software architecture are presented. The software framework employs seventeen basic reliability scenarios that serve as criteria for evaluating information trustworthiness, enabling the system to integrate seamlessly with diverse banking products (including the bank operating day environment). This ensures comprehensive testing, identification of abnormal transactions, and detection of potential abuses in banking practice. The developed crossplatform application is described, highlighting its modular structure, scalability, and suitability for implementation in contemporary financial monitoring systems. The architecture also allows for extension with additional analytical modules and supports the integration of external data sources, making the solution adaptable to evolving regulatory requirements and new patterns of financial risk.

Keywords: Bayesian logical-probabilistic model; financial monitoring; criterion algorithm; neural network model; fuzzy logic; crossplatform programming.