Temporal classification for fault-prediction in a real-world telecommunications network

Abstract

This paper presents a new temporal classification approach for fjldt-prediction in a Telecommutnications Network. The countrvwide data network of Pakistan Telecom (PTCL) has been selected as a basis for the investigation of classification algor-ithms to predict flults beJbre they stop a large numnber of users' circuits from normal operation. The main problenms addressed are the evalluation of alarms and clevelopment of newi, machine learning tools to help overcome the interoperabilitv issues. The motivation behind this work is to assist humian operators and nminimize the cost of the alarm evalulation