×

Analyzing and detecting anomalies in data records using artificial intelligence

  • US 7,689,455 B2
  • Filed: 05/16/2005
  • Issued: 03/30/2010
  • Est. Priority Date: 04/07/2005
  • Status: Active Grant
First Claim
Patent Images

1. A method comprising:

  • creating one or more context insight records representing an indication of a behavior of a service based on raw footprint data related to use of said service by a population of users;

    selecting at least one anomaly criteria dimension;

    defining possible values for multiple dimensions of said one or more context insight records in response to a result of a scanning of said one or more context insight records;

    analyzing, by an analyzer that is a hardware component, said one or more context insight records, in response to said at least one anomaly criteria dimension and to at least one of said possible values defined;

    detecting based on said analyzing a dynamically defined anomaly of said service by generating a first generation of candidate anomalies by segmenting said one or more context insight records along a segmentation dimension of said raw footprint data;

    generating a consecutive generation of candidate anomalies by modifying, adding or deleting one or more candidate anomalies of said first generation;

    said consecutive generation of candidate anomalies further comprising;

    modifying said first generation into said consecutive generation by activating a dimension of said one or more context insight records;

    selecting, by the hardware analyzing, at least one of each types of dimensions of said context insight records said dimensions comprising;

    criteria, segmenting, and testing dimensions;

    wherein said criteria dimensions represent characteristics of at least one context insight record population;

    wherein said segmenting dimensions divide said at least one context insight record population into several sub-populations;

    wherein said testing dimensions comprise criteria by which different sub-populations of the segmenting dimensions are tested;

    wherein the generating of at least one of the generations of candidate anomalies is response to, said criteria, segmenting and said testing dimensions selected by the hardware analyzer;

    repeating generating consecutive generations of candidate anomalies until a convergence condition holds true; and

    assisting in resolving a problem in the service in response to the dynamically defined anomaly.

View all claims
  • 5 Assignments
Timeline View
Assignment View
    ×
    ×