Parametric outlier detection
US7062415B2 · kind B2 · utility
Assignee
Inventors
Key dates
| Filing date | Aug 27, 2004 |
| Grant date | Jun 13, 2006 |
| Priority date | — |
| Expiry date | Sep 3, 2024 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG06T2207/30148
- WIPO fieldControl
- WIPO sectorInstruments
Abstract
A method for determining outlier data points in. A subset of dataset patterns is selected from a set of mathematical dataset patterns, and the subset of dataset patterns is combined into a composite dataset. The composite dataset is compared to the dataset, and a degree of correlation between the composite dataset and the dataset is determined. Data points within the composite dataset are selectively weighted to improve the degree of correlation, and the steps described above are selectively iteratively repeated until the degree of correlation is at least a desired value. Residuals for the data points within the composite dataset are selectively determined. At least one of the weighted data points within the composite dataset that are weighted within a first specified range, and data points within the composite dataset that have a residual within a second specified range, are selectively output as outlier data points.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.