Classification methods comparison for customer churn prediction in the telecommunication industry

Abstract

The need for telecommunication services has increased dramatically in schools, offices, entertainment, and other areas. On the other hand, the competition between telecommunication companies is getting tougher. Customer churn is one of the areas that each company gains more competitive advantage. This paper proposes a comparison of several classification methods to make a prediction whether the customers cancel the subscription to a telecommunication service by highlighting key factors of customer churn or not. The comparison is non-trivial due to the urgent requirements from the telecommunication industry to infer the most appropriate techniques in analyzing their customer churn. This comparison is often of huge commercial value. The result shows that Artificial Neural Network (ANN) can predict churn with an accuracy of 79%, Support Vector Machine (SVM) with 78% accuracy, Gaussian Naïve Bayes, and K-Nearest Neighbor (KNN) with 75% accuracy, while Decision Tree with 70% accuracy. Moreover, the technique with the highest F-Measure is Gaussian Naïve Bayes with 65% and the technique with the lowest one is Decision Tree with 49%. Hence, ANN and Gaussian Naïve Bayes are two methods with high recommendation to predict the customer churn in the telecommunication industry.

A Comparative Study on Customer Churn Analysis Using Machine Learning and Data Enrichment Techniques

“…After applying these algorithms to the unbalanced dataset, they were also used to balance the data with the SMOTE technique, and the results were compared. In another study, Makruf et al, [15] applied four different algorithms (ANN, SVM, K-Nearest Neighbors, Naive Bayes, and Decision Tree) to the customer data. After the application, the accuracy and F-scores of the algorithms were calculated and compared.…”

Multi-label text classification on unbalanced Twitter with monolingual model and hyperparameter optimization for hate speech and abusive language detection

“…The proliferation of labeled datasets has sparked considerable interest in classification across diverse domains, including telecommunications (Makruf et al, 2021), biology (Pratondo and Bramantoro, 2022), and law (Bramantoro and Virdyna, 2022). Several studies related to hate speech and abusive language have been proposed, and their datasets have been shared to support other researchers.…”

Leveraging LDA Feature Extraction to Augment Human Activity Recognition Accuracy

“…Specifically, Gaussian Naive Bayes is used when the features are continuous [95] and a particular type of Naive Bayes algorithm [96]. Its main advantage is that it requires a small amount of training data to determine the projected parameter needed for classification (Its characteristic of parameter independence may allow it to be used with fewer parameters than other techniques) [97].…”