Previsão de abandono para serviços de telecomunicações
Date
2024
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
IPCB. EST
Abstract
Este projeto foi desenvolvido com o intuito de criar um modelo de previsão de abandono de um serviço (customer churn) de telecomunicações utilizando algoritmos de Machine Learning (ML).
De modo a ser possível desenvolver um modelo de previsão de customer churn, foi necessário estudar as áreas de Inteligência Artificial, dedicadas a este tipo de previsão. Já existentes no setor da telecomunicação.
Elaborou-se um estudo do estado da arte do uso de técnicas de Machine Learning na previsão do customer churn no setor da telecomunicação. Este estudo evidenciou o crescente interesse atual neste tema e permitiu observar a capacidade dos modelos para fazerem previsões corretas.
Fez-se um estudo dos datasets utilizados para avaliar as distribuições, a sua representatividade e consistência.
Foram treinados 6 modelos de previsão e comparados entre si, fazendo ainda uma análise da importância de cada atributo dos modelos resultantes.
Adicionalmente para validar a generalização dos modelos, foram feitos alguns testes de previsão usando ao mesmo tempo dois datasets de contextos diferentes.
Desta maneira, a realização deste projeto permitiu identificar os melhores modelos de IA, com ênfase na capacidade de evitar a previsão de falsos casos de retenção.
Abstract : This project was developed with the aim of creating a model to predict the abandonment of a telecommunications service (customer churn) using Machine Learning algorithms. To be able to develop a predictive model for customer churn, it was necessary to study the areas of Artificial Intelligence, dedicated to this type of prediction that already exists in the telecommunication sector. To understand the work already done dedicated to solving this problem, a state-of-the-art study was carried out on the use of Machine Learning (ML) techniques in predicting customer churn in the telecom sector. This study highlighted the current growing interest in this topic and has shown the capacity of intelligent models to make correct predictions. A study was carried out on the selected datasets, to evaluate the distribution, their representativeness and consistency. Six prediction models were trained and compared with each other, and an analysis was also made of the importance of each attribute of the resulting models. In addition, to validate the generalization of the models, some prediction tests were carried out using two datasets from different contexts at the same time. In this way, the completion of this project made it possible to identify the best AI models, with an emphasis on the ability to avoid predicting false retention cases.
Abstract : This project was developed with the aim of creating a model to predict the abandonment of a telecommunications service (customer churn) using Machine Learning algorithms. To be able to develop a predictive model for customer churn, it was necessary to study the areas of Artificial Intelligence, dedicated to this type of prediction that already exists in the telecommunication sector. To understand the work already done dedicated to solving this problem, a state-of-the-art study was carried out on the use of Machine Learning (ML) techniques in predicting customer churn in the telecom sector. This study highlighted the current growing interest in this topic and has shown the capacity of intelligent models to make correct predictions. A study was carried out on the selected datasets, to evaluate the distribution, their representativeness and consistency. Six prediction models were trained and compared with each other, and an analysis was also made of the importance of each attribute of the resulting models. In addition, to validate the generalization of the models, some prediction tests were carried out using two datasets from different contexts at the same time. In this way, the completion of this project made it possible to identify the best AI models, with an emphasis on the ability to avoid predicting false retention cases.
Description
Keywords
Customer churn, Dataset, Machine learning, Previsão, Telecomunicações, Forecasting, Telecom
Citation
GONÇALVES, Francisco Mateus (2024) - Previsão de abandono para serviços de telecomunicações. Castelo Branco : IPCB. EST. Relatório do Trabalho de Fim de Curso de Engenharia Informática