User Feedback for Learning of Network-Incident Severity
A computer system that updates a pretrained predictive model is described. During operation, the computer system may receive, from an electronic device, information specifying user feedback about a network incident. Then, the computer system may update, based at least in part on the user feedback, the pretrained predictive model that outputs severity classifications of network incidents, where a difference between a severity classification of the updated pretrained predictive model and a user severity classification associated with the user feedback is reduced relative to an initial difference between an initial severity classification of the pretrained predictive model and the user severity classification. Moreover, the computer system may receive information specifying a second network incident. Next, the computer system may compute a second severity classification of the second network incident using the updated pretrained predictive model.
1 . A computer system, comprising:
an interface circuit configured to communicate with an electronic device;
a computation device electrically coupled to the interface circuit; and
memory, electrically coupled to the computation device, that stores program instructions, wherein, when executed by the computation device, the program instructions cause the computer system to perform operations comprising:
receiving, associated with the electronic device, information specifying user feedback about a network incident; and
updating, based at least in part on the user feedback, a pretrained predictive model configured to output severity classifications of network incidents, wherein a difference between a severity classification of the updated pretrained predictive model and a user severity classification associated with the user feedback is reduced relative to an initial difference between an initial severity classification of the pretrained predictive model and the user severity classification.
2 . The computer system of claim 1 , wherein the network incident comprises an event having an adverse impact on communication performance in at least a portion of a network.
3 . The computer system of claim 1 , wherein the operations comprise:
receiving information specifying a second network incident;
computing a second severity classification of the second network incident using the updated pretrained predictive model;
determining a priority on the second network incident based at least in part on the second severity classification; and
performing a remedial action based at least in part on the second severity classification, the priority or both.
4 . The computer system of claim 3 , wherein the remedial action comprises: changing a network configuration, correcting a component failure in the network, correcting a VLAN mismatch failure, or otherwise correcting the second network incident.
5 . The computer system of claim 3 , wherein the operations comprise computing a third severity classification of the second network incident using a general pretrained predictive model that is associated with multiple users, and the priority, the remedial action or both is based at least in part on the third severity classification.
6 . The computer system of claim 1 , wherein the pretrained predictive model, the updated pretrained model or both comprise: a neural network, a supervised machine-learning model, or another type of predictive model.
7 . The computer system of claim 1 , wherein the pretrained predictive model, the updated pretrained predictive model or both comprise a classifier or a regression model.
8 . The computer system of claim 1 , wherein the user feedback comprises: a score or a ranking corresponding to the user severity, a scope of the network incident, a type of the network incident, a physical location in the network of the network incident, or a hierarchical location in the network of the network incident.
9 . The computer system of claim 1 , wherein the computation device comprises a processor or a graphical processor unit.
10 . A non-transitory computer-readable storage medium for use in conjunction with a computer system, the computer-readable storage medium storing program instructions, wherein, when executed by the computer system, the program instructions cause the computer system to perform one or more operations comprising:
receiving, associated with the electronic device, information specifying user feedback about a network incident; and
updating, based at least in part on the user feedback, a pretrained predictive model configured to output severity classifications of network incidents, wherein a difference between a severity classification of the updated pretrained predictive model and a user severity classification associated with the user feedback is reduced relative to an initial difference between an initial severity classification of the pretrained predictive model and the user severity classification.
11 . The non-transitory computer-readable storage medium of claim 10 , wherein the network incident comprises an event having an adverse impact on communication performance in at least a portion of a network.
12 . The non-transitory computer-readable storage medium of claim 10 , wherein the operations comprise:
receiving information specifying a second network incident;
computing a second severity classification of the second network incident using the updated pretrained predictive model;
determining a priority on the second network incident based at least in part on the second severity classification; and
performing a remedial action based at least in part on the second severity classification, the priority or both.
13 . The non-transitory computer-readable storage medium of claim 12 , wherein the operations comprise computing a third severity classification of the second network incident using a general pretrained predictive model that is associated with multiple users, and the priority, the remedial action or both is based at least in part on the third severity classification.
14 . The non-transitory computer-readable storage medium of claim 10 , wherein the pretrained predictive model, the updated pretrained model or both comprise: a neural network, a supervised machine-learning model, or another type of predictive model.
15 . The non-transitory computer-readable storage medium of claim 10 , wherein the user feedback comprises: a score or a ranking corresponding to the user severity, a scope of the network incident, a type of the network incident, a physical location in the network of the network incident, or a hierarchical location in the network of the network incident.
16 . A method for updating a pretrained predictive model, comprising:
by a computer system:
receiving, associated with the electronic device, information specifying user feedback about a network incident; and
updating, based at least in part on the user feedback, the pretrained predictive model that outputs severity classifications of network incidents, wherein a difference between a severity classification of the updated pretrained predictive model and a user severity classification associated with the user feedback is reduced relative to an initial difference between an initial severity classification of the pretrained predictive model and the user severity classification.
17 . The method of claim 16 , wherein the method comprises:
receiving information specifying a second network incident;
computing a second severity classification of the second network incident using the updated pretrained predictive model;
determining a priority on the second network incident based at least in part on the second severity classification; and
performing a remedial action based at least in part on the second severity classification, the priority or both.
18 . The method of claim 17 , wherein the method comprises computing a third severity classification of the second network incident using a general pretrained predictive model that is associated with multiple users, and the priority, the remedial action or both is based at least in part on the third severity classification.
19 . The method of claim 16 , wherein the pretrained predictive model, the updated pretrained model or both comprise: a neural network, a supervised machine-learning model, or another type of predictive model.
20 . The method of claim 16 , wherein the user feedback comprises: a score or a ranking corresponding to the user severity, a scope of the network incident, a type of the network incident, a physical location in the network of the network incident, or a hierarchical location in the network of the network incident.