国产欧美精品一区二区,中文字幕专区在线亚洲,国产精品美女网站在线观看,艾秋果冻传媒2021精品,在线免费一区二区,久久久久久青草大香综合精品,日韩美aaa特级毛片,欧美成人精品午夜免费影视

To solve the problem that some outliers cannot be detected because they are covered by scattered feature space in multidimensional feature data, and alleviate the phenomenon that the results of current anomaly detection methods are poorly interpretable or not interpretable, a subspace-based multivariate anomaly detection algorithm with interpretability is proposed. Firstly, in the multi-dimensional feature space, the distribution test of the features of each dimension is carried out. On this foundation, a set of feature space is selected for each object, and then the outlier score is calculated for each object. In this process, the intermediate process products of the algorithm are efficiently used to interpret the algorithm results and improve users to understand the data. The experimental results show that it has good accuracy and running time, and well explains the abnormity of outliers.