System Error Prediction for Business Support Systems in Telecommunications Networks

Reliability and stability have been treated as the major requirements for the Business Support System (BSS) in telecommunications networks. It is crucial and essential for service providers to maintain good operating state of the BSS. In this article, we aim at system error prediction for a BSS, i.e., we predict occurrences of the abnormal state or behavior of the BSS. Because the occurrences of system errors are rare events in the BSS (i.e., the dataset of system status is highly imbalanced), it is highly challenging to use machine learning or deep learning algorithms to predict system error for the BSS. To address this challenge, we propose a machine learning-based framework for the system error prediction and a Frequency-based Feature Creation (FFC) algorithm to create new features to improve prediction. By adding the time-series information created by the existing features, the proposed FFC can amplify the effects of important features. Our experimental results show that the FFC significantly improves the prediction performance for the Random Forest algorithm.