Classifying data with Naive Bayes models

Even though you probably hear a lot about super-advanced ML methods such as deep learning, it's important to say that simpler methods have existed for years and have proven to be very efficient in many situations. Generally, it's always a good idea when you start with a data science problem to try out simpler models that have fewer parameters and are easier to tune. This will quickly give you a baseline to compare with more advanced techniques.

In this section, we'll review Naive Bayes models, a group of fast and simple classification algorithms.

Intuition

Naive Bayes models rely on Bayes' theorem, which defines an equation to describe the probability of an event, given the probability of related events. In the context of classification, it gives us an equation to describe the probability of a label, , given a set of features. In our handwritten digit recognition problem, this would translate to "the probability...