THE ROLE OF MACHINE LEARNING IN AUTOMATING COMPLEX DATABASE MIGRATION WORKFLOWS

Abstract

Database migration remains a critical yet complex task in enterprise modernization efforts, often involving heterogeneous data sources, legacy systems, and intricate schema transformations. Traditional approaches, which rely heavily on manual intervention and rule-based logic, are time-consuming, error-prone, and difficult to scale across large, distributed environments. This paper explores the application of machine learning (ML) techniques to automate key aspects of database migration workflows, including schema matching, data transformation, anomaly detection, and data quality validation. We propose a modular ML-driven framework that integrates supervised and unsupervised learning models to streamline end-to-end migration processes. A case study involving the migration of financial transaction data from Oracle to PostgreSQL demonstrates significant improvements in accuracy, reduction in manual effort, and consistency of schema alignment. Experimental results indicate that ML-enabled migration pipelines can reduce mapping errors by over 40% and decrease execution time by up to 60% compared to traditional methods. These findings highlight the potential of machine learning to transform legacy database migrations into more reliable, adaptive, and intelligent processes, enabling organizations to accelerate digital transformation while minimizing risk.