Towards a Unified Strategy for the Preprocessing Step in Data Mining