More the number of levels of a categorical variables more is the number of dummy variables we have to create and more difficult will it be to interpret the model. Re-grouping the levels of categorical variables rescues us from such a messed up model. For each levels of a categorical variables we will calculate the yes/no ratio of attrition and club the levels with similar yes/no ratio. The following table shows how we do it. The proportion columns in each table shows the values of yes/no proportion.