Manifold-anomaly detection with axis parallel explanations
AssigneeOptum Services (Ireland) Limited
InventorsLorcan B. Mac Manus, Conor Breen, Peter Cogan
Life of this patent
Estimating the term…
Abstract
Systems, methods, and apparatuses for detecting and identifying anomalous data in an input data set are provided.
Claims30
An apparatus for identifying anomalous data in an input data set comprising anomalous data, the apparatus comprising at least one processor and at least one memory comprising computer program code, the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus to:
- receive a plurality of data records from a data source, each data record of the plurality of data records comprising a feature vector comprising a plurality of predictor variables and a plurality of corresponding predictor variable values, wherein the plurality of data records comprises at least one anomalous data record;
- for each data record P of the plurality of data records,
- sequentially apply each ranked model of a plurality of ranked models to the data record P until a successful output is obtained from a ranked model;
- upon determining a successful output has been obtained from a ranked model, determine a predictor variable of the plurality of predictor variables that has been excluded from the ranked model from which the successful output was obtained;
- and
- identify a corresponding predictor variable value associated with the predictor variable as an anomalous value associated with the data record P.
A computer program product comprising a non-transitory computer readable storage medium storing computer code that, when executed by a computer, causes the computer to:
- receive a plurality of data records from a data source, each data record of the plurality of data records comprising a feature vector comprising a plurality of predictor variables and a plurality of corresponding predictor variable values, wherein the plurality of data records comprises at least one anomalous data record;
- for each data record P of the plurality of data records,
- sequentially apply each ranked model of a plurality of ranked models to the data record P until a successful output is obtained from a ranked model;
- upon determining a successful output has been obtained from a ranked model, determine a predictor variable of the plurality of predictor variables that has been excluded from the ranked model from which the successful output was obtained;
- and
- identify a corresponding predictor variable value associated with the predictor variable as an anomalous value associated with the data record P.
A computer implemented method, comprising:
- receiving a plurality of data records from a data source, each data record of the plurality of data records comprising a feature vector comprising a plurality of predictor variables and a plurality of corresponding predictor variable values, wherein the plurality of data records comprises at least one anomalous data record;
- for each data record P of the plurality of data records,
- sequentially applying each ranked model of a plurality of ranked models to the data record P until a successful output is obtained from a ranked model;
- upon determining a successful output has been obtained from a ranked model, determining a predictor variable of the plurality of predictor variables that has been excluded from the ranked model from which the successful output was obtained;
- and
- identifying a corresponding predictor variable value associated with the predictor variable as an anomalous value associated with the data record P.
Dependencies are inferred from the claim text.