The exponential growth of heterogeneous data from diverse sources, such as social media, IoT sensors, and transactional databases, poses significant challenges for effective processing and analysis. This data, often characterized by poor quality, diverse formats, and complex structures, prevents its utilization for extracting valuable insights and supporting informed decision-making. Machine learning (ML) emerges as a powerful tool to address these challenges by automating heterogeneous data processing tasks and enhancing data quality, integration, and analysis. In this context, this paper explores the contribution of machine learning methods to the different stages of the data management process: preparation, integration, and analytics. We aim to provide a comprehensive study of the role these methods play throughout the entire pipeline, as well as highlighting a set of challenges in this field.