Classification Problems

Classification problems are the most common type of machine learning problem. Classification tasks are different from regression tasks, in the sense that, in classification tasks, we predict a discrete class label, whereas in the case of regression, we predict continuous values. Another notable difference between classification problems and regression problems lies in the choice of performance metrics. With classification problems, accuracy is commonly chosen as a performance metric, while root mean square is quite common in the case of regression.

There are many important business use cases for classification problems where the dependent variable is not continuous, such as churn and fraud detection. In these cases, the response variable has only two values, that is, churn or not churn, and fraud or not fraud. For example, suppose we are studying whether a customer churns (y = 1) or doesn't churn (y = 0) after signing up for a mobile service contract. Then, the probability...