Code repository for the paper: 'Bogaert, M., & Delaere, L. (2023). Ensemble Methods in Customer Churn Prediction: A Comparative Analysis of the State-of-the-Art. Mathematics, 11(5), 1137.'
The code has been built in R (version 4.2.2) and R-studio. All important libraries are loaded in the first code chunk. If the required packages are not installed, make sure that you run install.packages first. For catboost follow the guidelines on https://catboost.ai/en/docs/installation/r-installation-binary-installation. For scmamp follow the guidelines on https://github.com/b0rxa/scmamp.
Since the code is a R-markdown file, it is advised that you open the file in R-studio.
In the paper there are 11 data sets of which 3 are proprietary and 8 publicly available. Since these propietary data sets are confidential, they cannot be shared. Hence, the first 3 data sets in the code should be discarded. The other 8 data sets are online available and can be downloaded from the following links:
- https://www.kaggle.com/becksddf/churn-in-telecoms-dataset/data#
- https://www.kaggle.com/shubh0799/churn-modelling
- https://kdd.org/kdd-cup/view/kdd-cup-2009/Data
- https://www.kaggle.com/barun2104/telecom-churn?select=telecom_churn.csv
- https://www.kaggle.com/blastchar/telco-customer-churn
- https://www.kaggle.com/datasets/mahreen/sato2015
- https://www.kaggle.com/datasets/jpacse/datasets-for-churn-telecom?select=cell2celltrain.csv
- www.sgi.com/tech/mlc/db
To make sure that everyhting runs smoothly, put the data sets in the same directory as the code file.
Since the paper benchmarks 33 classifiers across all 8 data sets, the code takes very long to run. Users that want to speed up the running time are advised to parallellise the code with foreach or run each data set in a different R window.
Bogaert, M., & Delaere, L. (2023). Ensemble Methods in Customer Churn Prediction: A Comparative Analysis of the State-of-the-Art. Mathematics, 11(5), 1137.