摘要: Prediction in streaming data is an important activity various branches of science such as sociology, economics and politics. Two major challenges offered by streams are (1) the underlying concept may change over time; (2) grow without limit so that it difficult to retain a long history raw data. Previous research has mainly focused on manipulating relatively recent The distinctive contribution this paper three folds. First, uses measure conceptual equivalence organize into concepts. Transition patterns among concepts can be learned from help prediction. Second, carries out prediction at two levels, general level predicting each oncoming specific instance’s class. Third, proposes system RePro incorporates reactive proactive mechanisms predict with efficacy efficiency. Experiments conducted compare representative existing methods benchmark sets represent diversified scenarios change. Empirical evidence offers inspiring insights suggests proposed methodology advisable solution for streams.