ARCHITECTURE OF AN INFORMATION-ANALYTICAL SYSTEM FOR ADAPTIVE BANKING SYSTEM MANAGEMENT

Abstract

Fragmentation of information sources, analytical models, and visualization tools complicates reproduction of the evidence underlying a managerial decision and increases the risk of using asynchronous data, outdated models, or opaque analytical signals. The study aims to develop a reference architecture of an information-analytical system for adaptive banking system management in which the path from a data source to a decision is organized as a governed evidence cycle. A six-layer functional architecture is proposed, complemented by four cross-cutting governance domains and three control gates: Data Gate, Model Gate, and Decision Gate. The functional layers cover acquisition and metadata, integration and data quality, a semantic and feature layer, an analytical and model layer, decision orchestration, and presentation and interaction. The scientific contribution consists in formalizing evidence continuity along the chain “source – data – transformation – model – signal – explanation – decision – outcome – feedback”, with versioning, lineage, explainability by design, fail-safe analytics, and mandatory human-in-the-loop control. The minimum data-quality gate is empirically verified on 84 monthly observations of the integral index of adaptive development of the banking system (IAR) and five functional components for 2019M01-2025M12. The mean and minimum completeness coefficients equal 1, the maximum reconciliation error is approximately 2.22×10^-16, and all records receive PASS status. Synthetic negative-control scenarios are treated separately to test CHECK/STOP responses. The architecture is platform-independent and supports reproducible, accountable, governed, and auditable analytical decision support across heterogeneous information sources and model services. Its practical value lies in reducing information and model risk without equating an automated signal with an automated managerial decision.