Introduction

In previous chapters, where an introduction to machine learning was covered, you were introduced to two broad categories of machine learning; supervised learning and unsupervised learning. Supervised learning can be further pided into two types of problem cases, regression and classification. In the last chapter, we covered regression problems. In this chapter, we will peek into the world of classification problems.

Take a look at the following Figure 3.1:

Figure 3.1: Overview of machine learning algorithms

Classification problems are the most prevalent use cases you will encounter in the real world. Unlike regression problems, where a real numbered value is predicted, classification problems deal with associating an example to a category. Classification use cases will take forms such as the following:

• Predicting whether a customer will buy the recommended product
• Identifying whether a credit transaction is fraudulent
• Determining whether a patient has a disease
• Analyzing images of animals and predicting whether the image is of a dog, cat, or panda
• Analyzing text reviews and capturing the underlying emotion such as happiness, anger, sorrow, or sarcasm

If you observe the preceding examples, there is a subtle difference between the first three and the last two. The first three revolve around binary decisions:

• Customers can either buy the product or not.
• Credit card transactions can be fraudulent or legitimate.
• Patients can be diagnosed as positive or negative for a disease.

Use cases that align with the preceding three genres where a binary decision is made are called binary classification problems. Unlike the first three, the last two associate an example with multiple classes or categories. Such problems are called multiclass classification problems. This chapter will deal with binary classification problems. Multiclass classification will be covered next in Chapter 4, Multiclass Classification with RandomForest.