Data Integration for Data Science: Solutions and Still Open Problems
摘要
This paper, accompanying a talk at the ADBIS 2025 Doctoral Consortium School, provides an overview of various data integration (DI) architectures - from virtual through physical to hybrid. It also presents an architecture for integrating stream data from robotic devices and introduces a novel concept for managing data source (DS) connectors. Additionally, the paper discusses selected current trends in applying machine learning to DI problems and outlines a few open research challenges. The insights presented in the talk and paper are drawn from our practical experience in data integration projects across the financial, IT, and intelligent farming sectors.