PREDICTING PT X’S FIXED BROADBAND SERVICE DISRUPTION USING DATA MINING METHODS

• Main Author: Umar Fathurrohman, Muhammad
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/47014
• Institution: Institut Teknologi Bandung

PT X is a nationwide company which provides fixed broadband internet service. From April 2019 to June 2019, fixed broadband service received more than 800 thousands of internet disruption reports. These reports can be grouped into two groups, which are individual and mass disruption reports as well as technical and non technical disruption reports. More than 600 thousands of fixed broadband service users out of 5 million active users experienced technical disruptions with internet disruptions in Region 2 consistently having the highest number of disruption reports. PT X wants to improve quality of services by reducing the number of disruption reports by detecting disruption earlier without waiting reports from users. Currently, PT X does not have the tools yet to anticipate fixed broadband service disruptions. This research discussed model development for predicting fixed broadband service disruption in PT X using network performance data and disruption reports data. The Cross-Industry Standard Process for Data Mining is used to develop the predictive model. Predictive model is built using random forest, logistic regression, support vector machine, and artificial neural network. It is known that artificial neural network performs as the best predictive model on test with 65,55% mean of accuracy and 54,29% mean of recall. This research uses python programming language to build predictive model and to build prototype. The prototype can be used to predict user’s service disruption using best performing predictive model. It is expected that this prototype could reduce the number of fixed broadband disruption reports received by PT X.