In the current multi anomaly parallel detection of heterogeneous cloud resource cross database, the processing of data characteristics is insufficient, resulting in low accuracy of detection results.