基于趋势符号聚合近似的卫星时序数据分类方法

作为在时间序列数据挖掘中广泛使用的主要符号化表示方法，符号聚合近似（SAX）使用段的平均值作为符号表示，由于无法区分具有不同趋势但具有相同平均值符号的不同时间序列，某些情况下可能会导致错误的分类。提出了一种改进的符号表示——趋势符号聚合近似（TrSAX），集成SAX与最小二乘法，用以描述时间序列的均值和斜率，并由此构建出BOTS分类器。此外，对卫星的模拟量遥测时序数据中的角度序列、转速序列、电流序列进行分析，并从UCR公开数据集中筛选出与3种序列类似的3个数据集进行分类实验验证。与应用了SAX和2个改进的SAX、经典的欧氏距离（ED）、动态时间规整（DTW）的1-NN分类方法进行对比，结果表明:提出的BOTS分类方法的分类错误率明显低于其他5种分类方法。 

Satellite time series data classification method based on trend symbolic aggregation approximation

As the main symbolic representation method widely used in time series data mining, the Symbolic Aggregation Approximation (SAX) uses the mean value of segments as the symbolic representation. Since it is impossible to distinguish different time series that have different trends but the same mean value, it may lead to incorrect classification. This paper presents an improved symbol representation-Trend Symbol Aggregation Approximation (TrSAX), which integrates SAX and least squares method to describe the mean and slope value of the time series, and constructs the BOTS classifier. In addition, this paper analyzes the angle sequence, rotation speed sequence, and current sequence in the satellite analog telemetry time series data, and selects three datasets similar to these three sequences from the UCR public dataset for classification experiment verification. They are compared with the 1-NN classification methods using SAX, two improved SAX, classic Euclidean Distance (ED) and Dynamic Time Warping (DTW). The results show that the classification error rate of the proposed BOTS classification method is significantly lower than the other five classification methods.