Abstract: A method comprises: by a controller to communicate with devices of a network, storing classifiers assigned to groups of the devices to identify device commonality for each group; associating, to the classifiers, historical probabilities of success with which an automated response executed by one or more of the devices remediates a device alarm event; when a device of the devices reports the device alarm event, identifying each classifier to which the device belongs, each historical probability for each classifier, and a rule with classifier thresholds for the automated response; determining to execute the automated response on the device by evaluating the rule using each historical probability and the classifier thresholds; after the automated response is executed, performing a service test supported across the devices, and monitoring test results from the service test; and after the service test, updating each historical probability using the test results.