基于滑动窗口的定性趋势分析方法及应用  

Method and application of qualitative trend analysis with sliding window

作　　者：高东[1] 马昕[1] 许欣[1] 张贝克[1] 

机构地区：[1]北京化工大学信息科学与技术学院自动化系,北京100029

出　　处：《计算机应用研究》2014年第5期1441-1444,共4页Application Research of Computers

基　　金：中央高校基本科研业务费资助项目(ZY1208)

摘　　要：传统定性趋势分析方法中,存在划分片段宽度难以自适应、依赖人为设定阈值、算法复杂等问题,针对这些问题,提出了一种新的基于滑动窗口的定性趋势分析方法。方法采用一个滑动的窗口,并对窗口内的数据进行线性拟合,根据拟合的情况扩大或缩小窗口,确定每个片段的最佳大小,将每个片段识别为上升、下降或不变,最终获取数据的定性趋势。在化工过程趋势分析应用的结果表明,该方法能够更为有效地提取、识别出趋势,具有较高的准确性和较低的复杂度,为之后的数据压缩、故障诊断等打下坚实基础。In the traditional qualitative trend analysis, the width of segments cannot be adaptive, the threshold is dependent on human experience, and the algorithm is complicate. Due to these shortcomings, this paper proposed a new qualitative trend analysis method with a sliding window. In the method, data in a sliding window was fitted linearly. According to the result of linear fitting, it expanded or reduced the sliding window to determine the best width of a segment. When all the segments were determined, they were identified as increasing, decreasing or steady. Finally it obtained the qualitative trend of data. Its ap- plication on chemical data trend analysis proves that the method can extract and identify qualitative trend effectively with higher accuracy and lower complexity. It can lay a solid foundation for data compression and fault diagnosis.

关 键 词：定性趋势分析 滑动窗口 自适应 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]