Table 1.
Literature review of papers on churn prediction in telecommunication.
Fig 1.
The flowchart of the proposed method.
Fig 2.
The clustering process of the training dataset.
Fig 3.
The clustering process, removing atomic clusters and removing duplicate clusters.
Fig 4.
Creation of the evolutionary algorithm search space.
Fig 5.
The chromosome representation in the proposed method.
Table 2.
The dataset X including 10 samples and three classes a, b, and c.
Table 3.
The predictions of classifier C for samples of dataset X.
Table 4.
The number of correct predictions in each class.
Table 5.
Features of dataset.
Table 6.
Tuned parameters of algorithms.
Table 7.
The confusion matrix.
Fig 6.
Illustrating the set of non-dominated solutions in different generations with respect to the two objectives, ‘accuracy’, and ‘diversity’.
Fig 7.
The final population of the optimization algorithm based on the two goals of accuracy and diversity.
Fig 8.
The set of non-dominated solutions in different generations with respect to two objectives diversity and imbalance accuracy.
Fig 9.
The final population of the optimization algorithm based on the two goals of imbalance accuracy and diversity.
Table 8.
Comparison of the proposed algorithms with classical classifiers.
Fig 10.
Comparison of the proposed models with classical classifiers.
Table 9.
Comparison of the proposed algorithms with other ensemble models.
Fig 11.
Comparison of the proposed models with other ensemble models.
Table 10.
Comparison of the proposed two models with the presented models in the literature.