Practical Predictive Analytics

Examining the cox regression output

Since we are essentially running an adapted logistic regression, the coefficients in a cox model are always in log form. To transform them to a likelihood ratio you need to take the exponent. This is also part of the summary output.

First, take a look at the value of the exponentiated coefficient, listed under the exp(coef) column. Any coefficients significantly greater than 1 indicate that the customer is more likely to churn than not. On the other hand, exp(coef) < 1 indicates that the variable is less likely to churn. The magnitude of the difference from 1 will indicate higher probabilities in one direction.

The model results indicate that gender (males), satisfaction, and monthly charges are the significant ...