Simplified customer churn model based on Weehyong Tok's "Telco Customer Churn" Azure Machine Learning Studio experiment.
The experiment contains a two-class boosted decision tree and a two-class random forest producing two ROC curves. Hyperparameters are built to push ROC curve far and to compare it with an H2O machine learning project, for real-life usage and better computing performance number of decision trees should be kept under 100. The experiment doesn't contain data transformation and begins at the data split (default of 50%). Experiment with the same dataset, using R and H2O: https://github.com/szavuly/Machine-Learning/blob/master/Telco%20Customer%20Churn%20Markdown.Rmd