Analyzing large data sets to find deviation patterns
View Patent ↗Operations, such as data processing operations, can be improved by applying clustering and statistical techniques to observed behaviors in the data processing operations.
1. A computer-implemented method for analyzing a behavior of different entities performing similar tasks, the method comprising:
observing outputs of multiple similar tasks with similar inputs performed by different entities;
identifying clusters of outputs based on similarities and dissimilarities between outputs; and
determining a normative behavior for the entities based on a size of the clusters, wherein larger size clusters define the normative behavior.
2. The method of claim 1 further comprising:
identifying possible abnormal behavior, based on smaller size clusters that deviate from the normative behavior.
3. The method of claim 1 further comprising:
identifying a split in normative behavior, based on a large cluster evolving over time into two large clusters.
4. The method of claim 1 further comprising:
identifying a merge in normative behavior, based on two large clusters evolving over time into a single large cluster.
5. The method of claim 1 further comprising:
identifying an evolution in normative behavior, based on a large cluster evolving over time into a single large cluster characterized by different behavior.
6. The method of claim 1 wherein the entities performing tasks are doctors treating patients.
7. The method of claim 1 wherein the entities performing tasks are operators processing documents.
8. The method of claim 1 wherein the entities performing tasks are entities making financial decisions.
9. The method of claim 1 wherein the entities performing tasks are entities performing activities as part of a supply chain.
10. The method of claim 1 wherein the entities performing tasks are hospitals treating patients.
11. The method of claim 1 wherein the entities performing tasks are organizations issuing bills for goods and services rendered.
12. The method of claim 1 wherein the entities performing tasks include adverse event reports issued corresponding to a medicine or medical procedure or medical device.
13. The method of claim 1 wherein the entities performing tasks are monitoring reports on an electricity smart grid.
14. A computer program product for analyzing a behavior of different entities performing similar tasks, the computer program product stored on a tangible computer-readable medium and including instructions that, when loaded into memory, cause a processor to carry out the steps of:
observing outputs of multiple similar tasks with similar inputs performed by different entities;
identifying clusters of outputs based on similarities and dissimilarities between outputs; and
determining a normative behavior for the entities based on a size of the clusters, wherein larger size clusters define the normative behavior.
15. The computer program product of claim 14 wherein the computer program product causes the processor to carry out the further step of:
identifying possible abnormal behavior, based on smaller size clusters that deviate from the normative behavior.
16. The computer program product of claim 14 wherein the computer program product causes the processor to carry out the further step of:
identifying a split in normative behavior, based on a large cluster evolving over time into two large clusters.
17. The computer program product of claim 14 wherein the computer program product causes the processor to carry out the further step of:
identifying a merge in normative behavior, based on two large clusters evolving over time into a single large cluster.
18. The computer program product of claim 14 wherein the computer program product causes the processor to carry out the further step of:
identifying an evolution in normative behavior, based on a large cluster evolving over time into a single large cluster characterized by different behavior.
19. A computer-implemented method for analyzing a behavior of different entities, the method comprising:
identifying multiple similar tasks with similar inputs performed by different entities;
identifying clusters of outputs based on similarities and dissimilarities between outputs; and
determining a normative behavior for the entities based on a size of the clusters, wherein larger size clusters define the normative behavior.