A Predictive Model of Early Alerts to Improve the Availability of Internet and Data Service of a Telecommunications Company Through the XGBoost Algorithm

The evolution of information and knowledge technologies has caused disruptive business change and new business models, which are based on agile and secure access to data. Therefore, telecommunications companies need to provide quality services that guarantee continuity, efficiency, and privacy in the transmission of information. This paper addresses the problem of a national telecommunications company, which has detected low rates of availability and intermittence in the data and internet service of its corporate clients, partly generated by the inadequate management of link monitoring, with alerts and corrective actions, which have caused effects to the company such as loss of customers, fines, increase in operating costs, breach of service agreements, decrease in trust and corporate image, etc. In this virtue, we have proposed a predictive model based on data mining and machine learning techniques, which identifies patterns in the historical information of the network equipment. This model can be used to: predict link behavior, detect outages, generate early warnings, and proactively support the company’s decision-making. The CRISP-DM methodology has been used for the development and evaluation of the predictive model XGBoost (evolved technique of Decision Trees) in Python Jupyter Notebook. The results are encouraging with an effectiveness of up to 95.5% prediction.