Desarrollo de una herramienta de monitoreo y análisis predictivo mediante una técnica de machine learning para redes móviles 4G LTE

Abstract

This project describes an automated monitoring and alerting system to manage the percentage of downlink Physical Resource Blocks (PRBs) usage and user traffic over a given number of base station radio access network cells. Using Python and different APIs as data analysis tools, the system analyzes real network conditions, extracting logs from a database, while continuously and uninterruptedly monitoring the status of PRBs and traffic in each cell, in order to efficiently determine the status of a large number of cells in a very short time. The solution generates visual alerts via a bar graph on a web dashboard when there is a high usage of Physical Resource Blocks (PRBs), and when a low usage of Physical Resource Blocks (PRBs) is detected. These alerts will allow telecom operators to make proactive decisions without having to wait up to weeks for a team to manually identify cell by cell. In addition, the project incorporates machine learning (ML) techniques and statistical models, using Meta's Prophet library to make predictions about the percentage of Physical Resource Blocks (PRBs) usage in the downlink of cells. These predictions allow anticipating the behavior of the network 30 days or more ahead, facilitating proactive planning and improving the response capacity to possible congestion or underutilization of resources.