MULTI-DRIFT PREDICTIVE MONITORING FOR EVOLVING INFORMATION SYSTEMS

Abstract

This article addresses predictive monitoring of information systems under conditions of multidimensional functional-state evolution. Unlike conventional monitoring approaches focused on isolated anomalies, failures, or statistical deviations in data streams, the proposed approach treats an information system as a multilayer dynamic object influenced by interacting drift processes. The study considers nine drift types relevant to modern software-intensive and cyberinfrastructure environments: configuration, topology, role, policy, architectural, contextual, semantic, goal, and security drift. It is shown that these drifts affect not only current system parameters but also the validity of monitoring, interpretation, and decision-making processes. The current state of the field is analyzed and the literature is shown to remain fragmented across concept drift detection, multivariate change detection, software architecture erosion analysis, ontology evolution, role and policy evolution, context-aware access control, and self-adaptive systems. To address this fragmentation, the paper proposes an integrated predictive monitoring model based on an extended system state vector and a Predictive Drift Index for early identification of hazardous evolution trajectories. The model combines statistical, multivariate, architectural, contextual, semantic, and security-aware perspectives within a unified framework. A validation protocol is proposed, together with a simulation experiment based on controlled injection of isolated and combined drifts into a nine-dimensional system-state representation. The simulation demonstrates that the integrated predictive index reacts more clearly to multi-drift escalation than isolated indicators and supports earlier identification of degraded, vulnerable, anomalous, and critical trajectories. The proposed approach provides a basis for intelligent monitoring of evolving information systems.