Algoritmo de predicción de abandono de clientes de servicios fijos en telecomunicaciones basado en un conjunto de datos de la empresa IBM

Abstract

The competition that exists between telecommunications companies produces multiple offers for users, therefore, these companies are forced to advance strategies that allow keep customers or better yet expand the number of users. This phenomenon does not allow short-term dynamic generation of an appropriate service portfolio, which is identified with the main needs of customers causing massive cancellations of services. Therefore, this paper presents an algorithm that predicts the dropout rate of customers of a telecommunications service, so that companies mitigate this phenomenon through from an analysis. This analysis is built using descriptive statistics tools that establish metrics for the companies that provide the service and with the information collected Influential patterns are identified to assess the results over a period of time. The design of the algorithm conformed to several models that are compared with other techniques. Rported in the state of the art to refine the results obtained, and thus, determine the percentage of abandoned customers in a fixed telecommunications service. This input is the basis for structure action plans and execute them in order to improve the service.