DETECTION METHOD AND APPARATUS FOR MULTIVARIATE KPI TIME SERIES, AND DEVICE AND STORAGE MEDIUM

摘要

A detection method and apparatus for a multivariate KPI time series, and a computer device and a computer-readable storage medium, which relate to the technical field of artificial intelligence. The method comprises: according to a preset segmentation policy and a multivariate KPI time series to be trained, determining a plurality of corresponding segments of sub-multivariate KPI time series to be trained (S101), wherein the sub-multivariate KPI time series to be trained are untagged; training a preset model according to the sub-multivariate KPI time series to be trained, so as to generate a corresponding deterministic model (S102); acquiring a multivariate KPI time series to be tested (S103); according to the multivariate KPI time series to be tested and the deterministic model, acquiring a reconstruction probability value of a numerical value to be tested in the multivariate KPI time series to be tested (S104); and according to a preset threshold and the reconstruction probability value of the numerical value to be tested, determining whether a moment corresponding to the numerical value to be tested is an anomaly point (S105). The deterministic model is generated by means of the untagged multivariate KPI time series, the robustness of the model is enhanced, and the detection accuracy is improved.