Abstract: A method and system that separate thermal and background interference noise measurements from interference measurements can determine periodicity of an event using active percentage values. The active percentage value of an interference event may be determined based on unsupervised machine learning on a per-Recording Output Period (ROP) basis. The active percentage of the interference event may be used with other characteristics to help classify the interference source causing the interference. After measuring RF energy at a base station in the network at predetermined intervals over a plurality of consecutive first of time periods, for each time period, a method arranges binned values of the measured energy into at least one cluster, determines whether each cluster of the at least one cluster represents noise or interference, and characterizes the interference, and the method indicates a source of the interference.