CPC G06F 11/079 (2013.01) [G06F 11/076 (2013.01); G06F 11/0757 (2013.01); G06F 11/0778 (2013.01); G06N 20/00 (2019.01)] | 18 Claims |
1. A method for identifying a problematic compute resource, comprising:
providing first features associated with first previously-executed compute resources as first training data to a machine learning model, at least one user having interacted with the first previously-executed compute resources during at least one debug session, the first training data being positively-labeled as representing problematic features;
providing second features associated with second previously-executed compute resources as second training data to the machine learning model, the at least one user having not interacted with the second previously-executed compute resources during the at least one debug session, the second training data being negatively-labeled as representing non-problematic features;
determining, based on the machine learning model, a score indicating a likelihood that a compute resource of a plurality of compute resources is problematic;
generating a dependency graph that comprises a plurality of nodes representing the plurality of compute resources;
identifying the compute resource represented by a particular node as being problematic based on the score for the compute resource being above a predetermined score threshold; and
providing the dependency graph for display, wherein the dependency graph provided for display emphasizes the compute resource represented by the particular node as being problematic by graphically distinguishing at least one of the particular node or an edge connecting the particular node to a connected node from at least one of other nodes or other edges.
|