Hewlett-Packard Development Company, L.p.

Evan Kriminger

Choudur K Lakshminarayan

Jose Carlos Principe

Alexander Singh Alvarado

An example method for anomaly detection in streaming data includes applying statistical analysis to streaming data in a sliding window. The method also includes extracting a feature. The method also includes determining class assignment for the feature using class conditional probability densities and a threshold.

Anomaly detection in streaming data

Regime change in streaming data is detected. The streaming data is sent to a plurality of modules. Each module in the plurality of modules produces an association measure. The association measure is a measure of similarity between the streaming data and regime data associated with the module producing the association measure. A regime change in the streaming data is detected based on values of the association measures from the plurality of modules.

Detecting regime change in streaming data

A system and method for classifying unclassified samples. The method includes detecting a number of classes including training samples in training data sets. The method includes, for each class, determining a vector for each training sample based on a specified number of nearest neighbor distances between the training sample and neighbor training samples, and determining a class distribution based on the vectors. The method also includes detecting an unclassified sample in a data set and, for each class, determining a vector for the unclassified sample based on the specified number of nearest neighbor distances between the unclassified sample and nearest neighbor training samples within the class, and determining a probability that the unclassified sample is a member of the class based on the vector and the class distribution. The method further includes classifying the unclassified sample based on the probabilities.

Classifying unclassified samples

Systems and methods for multi-regime detection in streaming data are disclosed. An example method includes generating vectors for each sample of the streaming data. The method also includes inducing mean independence of the vectors to find an embedded data trajectory. The method also includes comparing the embedded data trajectory with known data trajectories. The method also includes issuing an alert if the embedded data trajectory corresponds to a known data trajectory indicating an anomaly in the streaming data.

Growing community of inventors

Gainesville, FL, United States of America

Evan Kriminger