Mining data streams is a field of study that poses new challenges. This research delves into the study of applying different techniques of classification of data streams, and carries out a comparative analysis with a proposal based on similarity; introducing a new form of management of representative data models and policies of insertion and removal, advancing also in the design of appropriate estimators to improve classification performance and updating of the model. © 2011 Springer-Verlag.