中文摘要：

目前挖掘概念流动的数据流已经成了研究热点。文章提出了一种既能很好地处理概念漂移又能从单类别中学习的算法UP-AB。通过在超平面数据集和标准数据集上的实验,与PNB[1]算法比较,表明该算法具有更高的准确度,能更快地适应概念漂移。

英文摘要：

Section 1 of the full paper explains what we believe to be an effective adaptive Bayes（UP-AB） algorithm;its core consists of：（1） we use the nave Bayes classifier to classify the positive and unlabeled examples of data streams with concept drifts by calculating the conditional probabilities of negative examples;（2） when concepts are drifting,we modify their classification parameters by using the incremental Bayes algorithm.Section 2 utilizes the moving hyperplane data sets and two types of real data sets to compare the performance of our UP-AB algorithm with that of the PNB（positive nai¨ve Bayes） algorithm.The comparison results,presented in Figs.1 through 5,show preliminarily that our UP-AB algorithm has stronger adaptability and better classification precision than the PNB algorithm and that it can converge to target concepts with good precision and fast speed,indicating that our UP-AB algorithm is effective.