Churn prediction, or the task of identifying customers who are likely todiscontinue use of a service, is an important and lucrative concern of firms inmany different industries. As these firms collect an increasing amount oflarge-scale, heterogeneous data on the characteristics and behaviors ofcustomers, new methods become possible for predicting churn. In this paper, wepresent a unified analytic framework for detecting the early warning signs ofchurn, and assigning a "Churn Score" to each customer that indicates thelikelihood that the particular individual will churn within a predefined amountof time. This framework employs a brute force approach to feature engineering,then winnows the set of relevant attributes via feature selection, beforefeeding the final feature-set into a suite of supervised learning algorithms.Using several terabytes of data from a large mobile phone network, our methodidentifies several intuitive - and a few surprising - early warning signs ofchurn, and our best model predicts whether a subscriber will churn with 89.4%accuracy.
展开▼
