Artificial intelligence based self-organizing event-action management system for large-scale networks

At a rule processing unit of an evolving, self-organized machine learning-based resource management service, a rule of a first rule set is applied to a value of a first collected metric, resulting in the initiation of a first corrective action. A set of metadata indicating the metric value and the corrective action is transmitted to a repository, and is used as part of an input data set for a machine learning model trained to generate rule modification recommendations. In response to determining that the corrective actions did not meet a success criterion, an escalation message is transmitted to another rule processing unit.