IP Library Granted Patent US 12688083
Granted Patent B2
US 12688083 · App. 18/368,656 · Granted Jul 21, 2026

Intelligent recommendation of time series anomaly detection model pipelines

Inventors: Jun Wang (Xi'an, CN); Jing Xu (Xi'an, CN); Xiao Ming Ma (Xi'an, CN); Xue Ying Zhang (Xi'an, CN); Si Er Han (Xi'an, CN); Jing James Xu (Xi'an, CN); Wen Pei Yu (Xi'an, CN)
Assignee: International Business Machines Corporation
G06F11/079G06F11/0709
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12688083
App. No.
18/368,656
Granted
Jul 21, 2026
Kind
B2
Abstract

A time series anomaly detection method, system, and computer program product that processes time series data includes absorbing profiles of the time series data and anomaly types of a model as features, optimizing biased ranks to create optimized ranks through merging initial ranks with new ranks generated by real anomalies, and auto-suggesting the optimized ranks for saving a predetermined amount of data operation.

Claims (32)

1 . A computer-implemented time series anomaly detection method that processes time series data in order to provide improved model selection performance and reduced model bias, the method comprising:

generating initial ranks of a model utilizing synthesized anomalies in time series data;

profiling time series data and anomalies of a model as features;

generating new ranks with real anomalies in time series data;

optimizing biased ranks to create optimized ranks through merging the initial ranks with new ranks generated by real anomalies; and

auto-suggesting the optimized ranks for saving a predetermined amount of data operations, the optimized ranks used for selecting top N models, where N is an integer.

2 . The computer-implemented time series anomaly detection method of claim 1 , further comprising generating synthesized anomaly data by profiling the time series data of the model and the anomaly types.

3 . The computer-implemented time series anomaly detection method of claim 1 , wherein, during the optimizing, a relationship between the generated initial ranks and the profiled time series data of the model is stored.

4 . The computer-implemented time series anomaly detection method of claim 1 , wherein a value of N is set by a user.

5 . The computer-implemented time series anomaly detection method of claim 1 , embodied in a cloud-computing environment.

6 . A time series anomaly detection computer program product that processes time series data in order to provide improved model selection performance and reduced model bias, the time series anomaly detection computer program product comprising a computer-readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform:

generating initial ranks of a model utilizing synthesized anomalies in time series data;

profiling time series data and anomalies of a model as features;

generating new ranks with real anomalies in time series data;

optimizing biased ranks to create optimized ranks through merging the initial ranks with new ranks generated by real anomalies; and

auto-suggesting the optimized ranks for saving a predetermined amount of data operations, the optimized ranks used for selecting top N models, where N is an integer.

7 . The time series anomaly detection computer program product of claim 6 , further comprising generating synthesized anomaly data by profiling the time series data of the model and the anomaly types.

8 . The time series anomaly detection computer program product of claim 6 , wherein, during the optimizing, a relationship between the generated initial ranks and the profiled time series data of the model is stored.

9 . The time series anomaly detection computer program product of claim 6 , wherein a value of N is set by a user.

10 . A time series anomaly detection system that processes time series data in order to provide improved model selection performance and reduced model bias, the time series anomaly detection system comprising:

a processor; and

a memory, the memory storing instructions to cause the processor to perform:

generating initial ranks of a model utilizing synthesized anomalies in time series data;

profiling time series data and anomalies of a model as features;

generating new ranks with real anomalies in time series data;

optimizing biased ranks to create optimized ranks through merging the initial ranks with new ranks generated by real anomalies; and

auto-suggesting the optimized ranks for saving a predetermined amount of data operations, the optimized ranks used for selecting top N models, where N is an integer.

11 . The time series anomaly detection system of claim 10 , further comprising generating synthesized anomaly data by profiling the time series data of the model and the anomaly types.

12 . The time series anomaly detection system of claim 10 , wherein, during the optimizing, a relationship between the generated initial ranks and the profiled time series data of the model is stored.

13 . The time series anomaly detection system of claim 11 , wherein a value of N is set by a user.

14 . The time series anomaly detection system of claim 10 , embodied in a cloud-computing environment.

15 . The computer-implemented time series anomaly detection method of claim 1 , wherein ranks are numerical values calculated from basic metrics to measure model performance.