Crescent Entertainment has built a model to predict who is likely to continue their subscription and who will discontinue. They have used more granular data about viewing preferences, time spent viewing, time of day preference, day of week preference, location, to build their model. Although they do not have actual customer data on gender, income, education, they have some survey data from which they have been able to predict such personal characteristics from the variables that they do have data on$.$ They know about you than you may have told you. They have applied their model to a holdout sample of $10,000$ customers that had not been used for training. They then compared their model predictions to what these customers actually did. A confusion matrix showing the results is given below.