Journal of Engineering Science and Technology Vol. 15, No. 4 (2020) 2289 - 2300 © School of Engineering, Taylor’s University 2289 BROADBAND NETWORK FAULT PREDICTION USING COMPLEX EVENT PROCESSING AND PREDICTIVE ANALYTICS TECHNIQUES EMERSON RAJA J. 1, *, HOSSEN J., ERVINA E. M. N. 1 , TAWSIF K. 1 , JESMEEN M. Z. H. 1 1 Faculty of Engineering and Technology, Multimedia University, Melaka Campus, Jalan Ayer Keroh Lama, 75450, Melaka, Malaysia *Corresponding Author: emerson.raja@mmu.edu.my Abstract The customer satisfaction of the broadband network mostly depends on robustness of the service offered by Internet Service Providers (ISP). Providing uninterrupted network service is essential in this communication era even though interruption in internet connection is unavoidable. However, if it is predicted earlier, the consequences can be minimized. Hence, it is essential to accurately forecast the faults in internet connection for Telecom Companies. The proposed tool for predicting broadband network fault is made up of a combination of Complex Event Processing (CEP) and Predictive Analytics (PA) techniques. The PA is used to predict network faults using techniques such as Logistic Regression (LR) or Naïve Bayes (NB). CEP is used to perform the prediction in real-time on streaming events. In this paper the performance of predictive model configured with LR is compared with the one configured with NB. Both the models had been tested for its performance using appropriate data set received from telecommunication industry using precision-recall curve and accuracy. It was found that the prediction accuracy of LR model (89.65%) is better than that of NB model (86.25%). It was also noticed that the derived AUC of LR is 0.52 which is much higher than 0.21 of NB. Hence, it was concluded that the predictive model configured with LR is performing better than the one configured with NB. So, the proposed tool configured with LR model can be implemented for fault prediction in network management systems. Keywords: Complex event processing, Logistic regression, Machine learning, Naïve Bayes, Predictive analytics.