科研速览 · Science Skim继续刷下去 · Keep skimming →
◆ INTERNATIONAL JOURNAL OF COMPUTER ENGINEERING & TECHNOLOGY2026-07-31· Software deployment

A MACHINE LEARNING FRAMEWORK PROPOSAL FOR DEPLOYMENT RISK PREDICTION IN ENTERPRISE BPM PLATFORMS: A SIMULATION-BASED FEASIBILITY STUDY FOR FINANCIAL SERVICES CHANGE MANAGEMENT

Munisekhar Katta

原始摘要(英文原文)· Original abstract
Production deployments of enterprise Business Process Management (BPM)applications in regulated financial services environments carry significant operational risk: deployment failures can cause customer-facing disruptions, revenue loss, and regulatory scrutiny.Despite advances in CI/CD automation, deployment risk assessment on enterprise BPM platforms remains a largely manual, experiencedependent process that does not systematically use the telemetry already produced by modern deployment orchestration tools.This paper proposes DRPA (Deployment Risk Prediction Architecture), a machine learning framework designed to integrate with enterprise BPM deployment pipelines as an automated risk-scoring quality gate prior to production promotion.DRPA is designed to combine four signal dimensions: (1) structural complexity of the deployment artifact, (2) configuration drift between staging and production environments, (3) learned patterns from historical deployment outcomes, and (4) dependency-graph analysis of cascading impact across interconnected application components.We describe the proposed architecture in detail and report an illustrative, author-constructed simulation intended to test the internal logic of the scoring approach and surface design questions, informed by the author's operational experience administering multi-stage deployment pipelines in a commercial lending BPM environment.The simulation is not an empirical evaluation: it uses synthetic, hand-parameterized data rather than measured production telemetry, and its directional findings -that combining artifact-level, environment-level, historical, and dependency-graph signals appears more informative than any single dimension alone -should be treated as motivation for a real pilot study rather than as evidence of real-world performance.We describe the audit-traceability properties the design is intended to support for SOX-and PCI-DSS-relevant change management, and we lay out the empirical validation this proposal still requires.
读原文 · Read the paper ↗

AI 追问PRO

登录后使用 AI 追问

讨论区

登录后参与讨论

相关论文 · Related

A MACHINE LEARNING FRAMEWORK PROPOSAL FOR DEPLOYMENT RISK PREDICTION IN ENTERPRISE BPM PLATFORMS: A SIMULATION-BASED FEASIBILITY STUDY FOR FINANCIAL SERVICES CHANGE MANAGEMENT — 科研速览 Science Skim