基于快速动态时间规划的模糊聚类算法

Fuzzy clustering algorithms based on fast dynamic time warping

为快速准确地计算时间序列数据相似度,引入快速动态时间规划距离(fast dynamic time warping,FDTW),提出了基于FDTW的模糊C均值算法和模糊C中心点聚类算法。FDTW通过对数据序列进行拉伸和压缩匹配时间序列数据,只要形状相同,即使发生时间位移也可以准确识别,同时解决了传统DTW计算效率较低的问题。试验结果表明,提出的算法仍能保证聚类的精度。

In order to measure the similarity between pair-wise time series data rapidly and accurately,fast dynamic time warping（ FDTW） distance is introduced,and the fuzzy C Means algorithm and fuzzy C medoids algorithm are both improved. The FDTW distance,which synchronizes time series by stretching or compressing data and can measure time series accurately even if they are asynchronous,has great advantage over the DTW distance in improving the computational efficiency. Experimental results show that the proposed algorithms could achieve the clustering precision.