Detection of TC-type anomalies in time series based on a weighted MCD approach
Xisheng Huang, Zhijian Wang · 2025
Detection of time series outliers is a key issue in time series analysis, especially in practical applications, which often face the “masking effect”. In this paper, we propose an outlier detection method based on the minimum covariance determinant (MCD), and on this basis, we introduce time weights to more accurately identify the transient change (TC-type) outliers in time series. The method leverages MCD’s robustness and incorporates temporal weighting to distinguish local anomalies from random fluctuations. Through simulation experiments and empirical analysis, this paper verifies the advantages of the MCD method under different pollution rates and sample sizes, and the improved method shows good outlier identification. Finally, the superiority of the robust detection method is further validated in the detection of financial market anomalies, which provides a new solution for the detection of outliers in complex dynamic data environments.