Mining of High Average-Utility Alarm Rules in Telecommunication Network Data

Abstract

Mining huge number of alarm events data collected at the network operations center of a mobile network operator has become a challenging problem in the mobile communication area due to the fact that alarm events are heterogeneous and that they have different significance levels and may occur more than once in a certain time-window. In this study, the High Average-Utility Itemset Mining (HAUIM) approach is adopted to identify high average-utility itemsets in the alarm events data collected at the network operations center of Turkcell, which is a major mobile network operator in Turkiye. Moreover, a new interestingness measure has been proposed to obtain association rules between high average-utility itemsets. Experimental results have shown the efficiency of the proposed system with respect to compression and prediction performances. © 2023 IEEE.