Customer churn is a major problem for businesses in an environment where competitors are constantly endeavoring to retain their client base as well as wooing the business of new clients. Churn or attrition occurs when customers start viewing a competing company more favorably and cease to do business with their current company. It is often said that, depending on the type of industry, acquiring a new customer is anywhere from five to 25 times more expensive than retaining an existing one. Therefore it makes sense for businesses of all kinds to understand their customers' behaviours using data and take appropriate actions in the face of emerging patterns heralding the departure of a body of valuable customers.
In this project, we are going to predict customer churn using the Telecom Dataset hosted by IBM Watson Analytics Community. Telecommunications is an industry in which subscribers have multiple competing companies to choose from over their current company. We will be predicting churn using a variety of machine learning models. In keeping with our philosophy of trying the simplest approaches first, we will start with linear models such as Logistic Regression, Linear Discriminant Analysis, and Support Vector Machines with linear kernel. We will then move on to non-linear models including Decision Trees and Neural Networks and culminate with some of the more complex models namely Gradient Boosting, Random Forests, and AdaBoost.