Mining data streams has attracted the attention of the scientific community in recent years with the development of new algorithms for processing and sorting data in this area. Incremental learning techniques have been used extensively in these issues. A major challenge posed by data streams is that their underlying concepts can change over time. This research delves into the study of applying different techniques of classification for data streams, with a proposal based on similarity including a new methodology for detect and treatment of concept change. Previous experimentation are conduced with the model because it have some parameters to be tuned. A comparative statistical analysis are presented, that shows the performance of the proposed algorithm. © 2012 IEEE.