【摘要】 时序降维是解决时间序列高维问题的关键技术。符号聚集近似表示(SAX表示法)作为一种时序降维技术,具有良好的维度约简能力与性能稳定的下界距离算法,但算法中分段数的选取需根据当前时序数据的特征而人为设定。针对这一问题,引入了滑动窗口算法与统计学方法,提出了基于二分迭代SAX的时序相似性度量算法。实验结果表明,该算法不仅解决了分段数设定困难的问题,而且降低了时序降维表示的复杂度,提高了SAX算法在多种时序数据上的分类准确性。更多还原

【Abstract】 Time series dimentionality reduction technology is used to resolve high dimensionality time series.Symbolic aggregate appro ximation(SAX representation)is a time series dimensionality reduction technique which benefits from its brief representation in dimensionality reduction and high-performance lower bound distance algorithm,but there is a question that the number of segments,aparameter in SAX,is set artificially based on the characteristic of individual time series.To solve this problem,similarity measure algorithm of time series based on binary-dividing SAX was presented by introducing sliding window and statistical methods.The experimental results show that binary-dividing SAX algorithm not only solves the difficulty to choose the number of segments,but also reduces the complexity of time series representation in dimensionality reduction and improves classification accuracy by using the SAX algorithm in a variety of time series data.更多还原