Machine Learning in a Changing World

For many different fields of research the target of the research evolves over time. In this paper a machine learning approach called DynamicWEB is presented. This allows a knowledge hierarchy to evolve as the result of changes in multiple observations of the same entities. By adapting the hierarchy to these changes, the learning technique is able to preserve and improve itself instead of being affected negatively by it. Several machine learning datasets are examined that demonstrate the value of this technique.