基于形变异常度的函数型数据异常值检测算法改进

Since the functional data outlier detection algorithm based on modified band depth and modified epigraph index is fail to detect the curves which is just abnormal in a small range, we propose a new shape-outlyingness and improve the outlier detection algorithm in order to solve this problem. First, we calculate the data depth of the functional data’s derivative, eliminate the abnormal curve whose depth value is small, then compute the trimming mean curve, and calculate the shape-outlyingness which is the degree of deviation between the derivative curve and the average curve, Finally, we modify the outlier detection algorithm based on shape-outlyingness, which supply the functional data’s shape information. The simulation results show that the improved outlier detection algorithm can effectively identify the curves whose anomaly rage is small, and it also improves the shape outliers’ detection rates.