Churn is a one of the biggest problem in the telecom industry. Research has shown that the average monthly churn rate among the top 4 wireless carriers in the US is 1.9% - 2%. In this analysis I have first developed some hypothesis by conducting a detailed exploratory data analysis for churn in the telecom industry and used predictive analytics to accurately identify the customers who are likely to churn.
I have explored the following classficiation techniques and compared their accuracy and other metrics such as ROC-AUC score, True Positive Rate and False Positive Rate.
Models Explored -
- Logistic Regression
- Random Forests
- Supprt Vector Machines (SVMs)
- ADABoost
- XG Boost
Overall approach is as follows -
- Data Preprocessing - To check for any missing values, correlated variables and dummy encoding
- EDA to develop hypothesis
- Scaling data to make sure we are using scaled data to improve accuracy
- Splitting data into train & test datasets
- Training each of the above models using cross validation
- Testing the accuracy on test data and plotting ROC curves for each of the models
- Comparing each model on various metrics and identifying the important parameters which help explain churn.