Benchmark test method and device for anomaly detection algorithm

The embodiment of the invention provides a benchmark test method of an anomaly detection algorithm, which is applied to anomaly detection of time series data, and comprises the steps of obtaining a set composed of a plurality of anomaly detection algorithms to be evaluated and a time series data set, the time series data set comprising a plurality of pieces of time series data; training each to-be-evaluated anomaly detection algorithm in a set through a training normal form and data in a training data set to obtain algorithm instances corresponding to each to-be-evaluated anomaly detection algorithm, the training set being a subset of the time sequence data set; and through each current algorithm instance, testing the to-be-evaluated anomaly detection algorithm corresponding to the current algorithm instance, and determining the detection effect of each current anomaly detection algorithm. According to the invention, benchmark testing can be carried out for different anomaly detection algorithms, and the perf