In various embodiments, a prediction subsystem automatically predicts a level of network availability of a device network. The prediction subsystem computes a set of predicted attribute values for a set of devices attributes associated with the device network based on a trained recurrent neural network (RNN) and set(s) of past attribute values for the set of device attributes. The prediction subsystem then performs classification operation(s) based on the set of predicted attribute values and one or more machine-learned classification criteria. The result of the classification operation(s) is a network availability data point that predicts a level of network availability of the device network. Preemptive action(s) are subsequently performed on the device network based on the network availability data point. By performing the preemptive action(s), the amount of time during which network availability is below a given level can be substantially reduced compared to prior art, reactive approaches.
BACKGROUND Field of the Various Embodiments (1) Embodiments of the present invention relate generally to networking technology and, more specifically, to techniques for analyzing a network to increase network availability. Description of the Related Art (2) Ensuring that network devices are available to users is an important aspect of providing effective network-based services. Typically, if the operation of one or more devices included in a device network is deficient, then the overall availability of the device network can be compromised, and some users may be unable to access or use the services provided by the device network. To maintain appropriate levels of network availability, service providers oftentimes implement monitoring tools that automatically monitor a variety of attributes related to the operation of different devices included in a device network. If, at any point in time, a monitoring tool receives an attribute value that indicates that the operation of a particular network device is deficient, then the monitoring tool generates an alert. When an engineer notices the alert, the engineer is able to troubleshoot the device and perform remediation operations on the device and/or device network to restore network availability to the appropriate level. (3) One drawback of the “reactive” approach to network analysis and remediation described above is that the operation of at least one network device needs to be deficient before a monitoring tool generates an alert
1. A computer-implemented method, comprising: for each device included in a plurality of interconnected devices within a device network, executing a trained recurrent neural network (RNN) that generates a set of predicted attribute values for a set of device attributes associated with the device based on input that includes at least one set of past attribute values for the set of device attributes, wherein the set of device attributes characterizes an operation of the device within the device network, and wherein the set of predicted attribute values is associated with a forward-looking time step and the at least one set of past attribute values is associated with one or more time steps occurring prior to the forward-looking time step; and performing one or more classification operations based on the sets of predicted attribute values for the plurality of interconnected devices and one or more machine-learned classification criteria to predict a plurality of probabilities for a plurality of levels of degradation in a network availability of the device network, wherein at least one preemptive action is subsequently performed on the device network based on the plurality of probabilities for the plurality of levels of degradation in the network availability. ||
10. One or more non-transitory computer readable media including instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of: for each device included in a plurality of interconnected devices within a device network, executing a trained recurrent neural network (RNN) that generates a set of predicted attribute values for a set of device attributes associated with the device based on input that includes at least one set of past attribute values for the set of device attributes, wherein the set of device attributes characterizes an operation of the device within the device network, and wherein the set of predicted attribute values is associated with a forward-looking time step and the at least one set of past attribute values is associated with one or more time steps occurring prior to the forward-looking time step; and performing one or more classification operations based on the sets of predicted attribute values for the plurality of interconnected devices and one or more machine-learned classification criteria to predict a plurality of probabilities for a plurality of levels of degradation in a network availability of the device network, wherein at least one preemptive action is subsequently performed on the device network based on the plurality of probabilities for the plurality of levels of degradation in the network availability. ||
19. A system, comprising: one or more memories storing instructions; and one or more processors that are coupled to the one or more memories and, when executing the instructions, are configured to: for each device included in a plurality of interconnected devices within a device network, execute a trained recurrent neural network (RNN) that generates a set of predicted attribute values for a set of device attributes associated with the device based on input that includes at least one set of past attribute values for the set of device attributes, wherein the set of device attributes characterizes an operation of the device within the device network, and wherein the set of predicted attribute values is associated with a forward-looking time step and the at least one set of past attribute values is associated with one or more time steps occurring prior to the forward-looking time step; and perform one or more classification operations based on the sets of predicted attribute values for the plurality of interconnected devices and one or more machine-learned classification criteria to predict a plurality of probabilities for a plurality of levels of degradation in a network availability of the device network, wherein at least one preemptive action is subsequently performed on the device network based on the plurality of probabilities for the plurality of levels of degradation in the network availability.