<abstract xmlns="http://www.w3.org/1999/xhtml"><p>The telecommunications industry is representative when it comes to a country’s economy. In this industry, the customer plays a very important role in maintaining a stable income. The churn customer is one of the most important concerns for large companies. This increased attention is due to its direct effect on the revenues of large companies in the telecommunications industry, companies being in a constant search to develop ways to predict this type of customer. The aim of our paper is to identify potential customers at risk of churn using modern data mining techniques, often used in the business world. From the nine techniques tested, we choose as the churn prediction model, the technique with the highest performance. The effectiveness of the model is tested and evaluated by the f1-score. The model developed in the paper uses machine learning techniques on the Python platform, exploring a wide range of algorithms from logistic regression and the method of balancing the analyzed data set (Balanced Random Forest) to supervised learning methods (K-Nearest Neighbors, Naive Bayes) and optimization packages (Ligh GBM, CATBoost, ADABoost, RUSBoost, Stochastic Gradient Descent). The techniques analyzed in this paper cover a diverse range of methods that are compared in terms of performance. RUSBoost proves to be the best churn prediction model for telecom customers in this study. RUSBoost has the lowest loss function of all the tested techniques.</p></abstract>

The telecommunications industry is representative when it comes to a country’s economy. In this industry, the customer plays a very important role in maintaining a stable income. The churn customer is one of the most important concerns for large companies. This increased attention is due to its direct effect on the revenues of large companies in the telecommunications industry, companies being in a constant search to develop ways to predict this type of customer. The aim of our paper is to identify potential customers at risk of churn using modern data mining techniques, often used in the business world. From the nine techniques tested, we choose as the churn prediction model, the technique with the highest performance. The effectiveness of the model is tested and evaluated by the f1-score. The model developed in the paper uses machine learning techniques on the Python platform, exploring a wide range of algorithms from logistic regression and the method of balancing the analyzed data set (Balanced Random Forest) to supervised learning methods (K-Nearest Neighbors, Naive Bayes) and optimization packages (Ligh GBM, CATBoost, ADABoost, RUSBoost, Stochastic Gradient Descent). The techniques analyzed in this paper cover a diverse range of methods that are compared in terms of performance. RUSBoost proves to be the best churn prediction model for telecom customers in this study. RUSBoost has the lowest loss function of all the tested techniques.

Churn prepaid customers classified by HyperOpt techniques

Proceedings of the International Conference on Applied Statistics

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

{"article-title":"Churn prepaid customers classified by HyperOpt techniques"}

The telecommunications industry is representative when it comes to a country’s economy. In this industry, the customer plays a very important role in...

Churn prepaid customers classified by HyperOpt techniques

Publicado en línea: 31 may 2021

Páginas: 107 - 119

DOI: https://doi.org/10.2478/icas-2021-0010

Palabras clave
Churn, customer, telecommunications, HyperOpt, F1-score

© 2021 Andreea Dumitrache et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Churn prepaid customers classified by HyperOpt techniques

Publicado en línea: 31 may 2021

Páginas: 107 - 119

DOI: https://doi.org/10.2478/icas-2021-0010

Palabras claveChurn, customer, telecommunications, HyperOpt, F1-score

© 2021 Andreea Dumitrache et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Palabras clave
Churn, customer, telecommunications, HyperOpt, F1-score