US Patent No. 10,599,543

UNSUPERVISED METHOD FOR BASELINING AND ANOMALY DETECTION IN TIME-SERIES DATA FOR ENTERPRISE SYSTEMS


Patent No. 10,599,543
Issue Date March 24, 2020
Title Unsupervised Method For Baselining And Anomaly Detection In Time-series Data For Enterprise Systems
Inventorship Sampanna Shahaji Salunke, Dublin, CA (US)
Dustin Garvey, Oakland, CA (US)
Uri Shaft, Fremont, CA (US)
Maria Kaval, Redwood Shores, CA (US)
Assignee Oracle International Corporation, Redwood Shores, CA (US)

Claim of US Patent No. 10,599,543

1. A method comprising:receiving a set of time-series data that includes a sequence of values captured by one or more computing devices over time;
detecting two or more seasonal patterns within the set of time-series data, including a first seasonal pattern and a second seasonal pattern;
generating, based on the set of time-series data, a first interval for the first seasonal pattern and a second interval for the second seasonal pattern;
wherein the first interval represents a first distribution of sample values associated with the first seasonal pattern;
wherein the second interval represents a second distribution of sample values associated with the second seasonal pattern;
monitoring a first set of one or more data points occurring at a first time period within a time-series signal for anomalies in the first seasonal pattern based, at least in part, on whether the first set of one or more data points falls outside of the first interval;
detecting, after the first time period, a transition between a first season when the first seasonal pattern is expected and a second season when the second seasonal pattern is expected;
responsive to detecting the transition, transitioning monitoring of the time-series signal from using the first interval to using the second interval such that a second set of one or more data points occurring at a second time period within the time-series signal are monitored for anomalies in the second seasonal pattern based, at least in part, on whether the second set of one or more data points falls outside of the second interval; and
in response to detecting an anomaly in the first seasonal pattern or the second seasonal pattern, generating an alert.