Abstract: In the current research, we present a supervised classification methodology with Light Gradient Boosting Machine (LightGBM) and Random Forest (RF) to predict the broadband telecommunication network customers' trouble tickets caused by network failures. Efficient fault diagnostics is of great concern for service providers when it comes to network management. Preventive or proactive maintenance measures are good in reducing the number of faults in the network, thus improving its availability and mitigation of network failures causing service interruption. Such measures help improve preventive maintenance, operational network maintenance, and workforce collaboration. The model in the current study is based on three datasets: Customer Trouble Tickets (CTT), Network Alarm data (NA), and Customer Premise Equipment (CPE). The forecasted model used historical trouble tickets on a Broadband Service provider in Kingdom of Bahrain, to predict future failures. In each phase, data sets are filtered according to specific criteria based on which predictive models are built using LightGBM and RF methods. RF is used to identify the most critical features. The predictive analysis results demonstrate that RF performs with an accuracy of 0.8442 which is the highest in comparison to other classification methods used in the current study. Therefore, we will choose the RF as our final model to predict the Alarm types. Keywords: Fixed Bahrain Broadband, Random Forest, LightGBM, Customer Experience, Network Fault Prediction