Customer churn reduction is a business scenario in which a company is trying to retain a customer which is more likely to leave the services. For reducing churn rate, we need to identify which customers are most likely to churn and which are not.
The solution is divided into 3 parts.
- Exploratory data analysis(EDA) was performed to explore the structure of data. Some of the basic assumptions were made about the data ie. Which variables are most likely causing churn. During exploration dataset was checked for missing values, multi collinearity and other model/algorithm specific assumptions.
- After EDA, for learning two models were used, logistic regression and random forest. Some data pre-processing was done to prepare training data for learning model.
- In the last part, performance tuning was done to increase the accuracy of models. Both the algorithms and EDA were implemented in R and python. Both implementations were similar with little difference due to difference in learning algorithm implementation.