There are many problems where the main objective is to develop a classification model that uses data to classify observations into two or more categories. For example, a patient may be classified as normal, suspect, or pathological based on the data on several variables. The deep learning network in this case will use data on several patients where the outcome is already available, and it will learn to classify a patient into one of the three categories.

Another example of a classification problem could be where students send applications to a graduate school. An application from a student may be accepted or rejected based on variables such as GPA, GRE, and ranking of the school during their undergraduate degree. Another interesting example could be where student-related data is used for developing a model that helps to classify first-year students into those that are likely to stay with the current school and those who are likely to transfer to another school. A similar model can be developed to classify customers who are likely to stay with a business or switch to a competitor. 

One of the challenges involved while developing a classification model is that of class imbalance. For example, when dealing with medical data, the number of patients classified as normal may be much larger than the number of patients who are classified as pathological. Similarly, when applying to a graduate program at one of the top universities, it is very likely that the data contains a significantly higher number of cases where an applicant is not accepted. Deep network models are useful in addressing such concerns easily. The Keras library used in this book provides a user-friendly interface not only to address such issues easily, but also to help in obtaining suitable classification models with the help of fast experimentation.

In Chapter 2, Deep Neural Networks for Multi-Class Classification, we provide an illustration of a multi-class deep learning classification model using R.