It is very likely that most information systems have at short-term information management problems. This requires the creation of new types of data management techniques more efficient and specific to each case, with the capacity to govern and ensure compliance with the management measures defined for operational systems, and ensure the desired performance and quality. In this work, we address the problem of data management, and, using a solution based on machine learning techniques, we tried to perceive, learn and classify the data contained in any database, according to its relevance for the users. Being able to identify what is really important to the users and separate this information from the rest, it is a great way for reducing the size of unnecessary data in a system and to define a more appropriate management model for the data that must be maintained in the system.