Abstract
Systems and methods for data classification and novel data class detection are provided. In one illustrative embodiment, a system or method for detecting a novel class includes receiving a data stream comprising a plurality of data points, and identifying a set of filtered outliers, in the plurality of data points, that are outside of a decision boundary. A cohesion and a separation for the set of filtered outliers may be determined. A novel class may be detected using the cohesion and the separation of the set of filtered outliers, and the novel class may include the set of filtered outliers.
| Original language | English (US) |
|---|---|
| U.S. patent number | 9165051 |
| Filing date | 8/22/11 |
| State | Published - Oct 20 2015 |