Comparative Study of Customer Churn Prediction Based on Data Ensemble Approach

Customer Churn is the prima facie problem of any company irrespective of their size and also it has a direct influence on the revenue of the companies. So this makes every company to understand the rationale behind the customer churn and predict the potential customer churn and the reasons for the same. By identifying the factors of churn, the companies will be in adept position to retain the customers. Thence the main intention of this work is to build-up a churn prediction model using machine learning that assists the companies to predict the customers who are most likely to churn. The models were measured by adopting Area Under Curve (AUC) measure, the confusion matrix (Accuracy, Recall, Precision) and F1 Score. The ensemble modeling was also adopted in order to compare and consolidate the each models for a common outcome. The models were run through Python (Jupyter Notebook) environment on the big dataset obtained from Kaggle. This research tried out with following model algorithms: Naive Bayes Model, Discriminant Analysis, Decision Tree, Random Forest, and Logistic Regression and further to attest the outcome of the data ensemble modeling of Ada boost, Gradient Boosting and Bagging has also been used. Out of all the models Random Forest model was considered to be the best fit model due to high accuracy, specificity, AUC and FI Score. So this model can be used as a classification churn prediction model.