关键词: 分类, 神经网络, 数据流, 可变滑动窗口, 趋势分析

Abstract: To improve the accuracy and adaptability of the classification of data stream, this paper presented a new method of classification. This method used total least squares to fit the segmentation of data stream, and presented a variable sliding window algorithm to achieve a reasonable segmentation and improve the accuracy of trend analysis by combining Sliding Window (SW) algorithm with extrapolation for On-line Segmentation of Data (OSD) algorithm. By extracting the dynamic feature of data stream, neural network was used for the pattern recognition of data stream and classification so it can provide early warnings, severity assessments of monitored subjects and information for decision support. The test results show that this method can describe the dynamic characteristic of data steam effectively, and the effect of classification is evident.

Key words: Classify, Neural network, Data stream, Changeable sliding window, Trend analysis